🔭Data Science: Metrics (Just the Quotes)

"Little experience is sufficient to show that the traditional machinery of statistical processes is wholly unsuited to the needs of practical research. Not only does it take a cannon to shoot a sparrow, but it misses the sparrow! The elaborate mechanism built on the theory of infinitely large samples is not accurate enough for simple laboratory data. Only by systematically tackling small sample problems on their metrics does it seem possible to apply accurate tests to practical data." (Ronald A Fisher,Statistical Methods for Research Workers", 1925), 

"The degree of confirmation assigned to any given hypothesis is sensitive to properties of the entire belief system [...] simplicity, plausibility, and conservatism are properties that theories have in virtue of their relation to the whole structure of scientific beliefs taken collectively. A measure of conservatism or simplicity would be a metric over global properties of belief systems." (Jerry Fodor,Modularity of Mind", 1983)

"No subjective metric can escape strategic gaming [...] The possibility of mischief is bottomless. Fighting ratings is fruitless, as they satisfy a very human need. If one scheme is beaten down, another will take its place and wear its flaws. Big Data just deepens the danger. The more complex the rating formulas, the more numerous the opportunities there are to dress up the numbers. The larger the data sets, the harder it is to audit them." (Kaiser Fung,Numbersense: How To Use Big Data To Your Advantage", 2013)

"The urge to tinker with a formula is a hunger that keeps coming back. Tinkering almost always leads to more complexity. The more complicated the metric, the harder it is for users to learn how to affect the metric, and the less likely it is to improve it." (Kaiser Fung," Numbersense: How To Use Big Data To Your Advantage", 2013)

"Precision and recall are ways of monitoring the power of the machine learning implementation. Precision is a metric that monitors the percentage of true positives. […] Recall is the ratio of true positives to true positive plus false negatives." (Matthew Kirk,"Thoughtful Machine Learning", 2015)

"Just as with aggregated data, an average is a summary statistic that can tell you something about the data - but it is only one metric, and oftentimes a deceiving one at that. By taking all of the data and boiling it down to one value, an average" (and other summary statistics) may imply that all of the underlying data is the same, even when it’s not." (John H Johnson & Mike Gluck,Everydata: The misinformation hidden in the little data you consume every day", 2016)

"Roughly stated, the No Free Lunch theorem states that in the lack of prior knowledge" (i.e. inductive bias) on average all predictive algorithms that search for the minimum classification error" (or extremum over any risk metric) have identical performance according to any measure." (N D Lewis,Deep Learning Made Easy with R: A Gentle Introduction for Data Science", 2016)

"Variables which follow symmetric, bell-shaped distributions tend to be nice as features in models. They show substantial variation, so they can be used to discriminate between things, but not over such a wide range that outliers are overwhelming." (Steven S Skiena,The Data Science Design Manual", 2017)

"Dimensionality reduction is a way of reducing a large number of different measures into a smaller set of metrics. The intent is that the reduced metrics are a simpler description of the complex space that retains most of the meaning." (Danyel Fisher & Miriah Meyer,Making Data Visual", 2018)

"Statistical metrics can show us facts and trends that would be impossible to see in any other way, but often they’re used as a substitute for relevant experience, by managers or politicians without specific expertise or a close-up view." (Tim Harford,The Data Detective: Ten easy rules to make sense of statistics", 2020)

"Non-linear associations are also quantifiable. Even linear regression can be used to model some non-linear relationships. This is possible because linear regression has to be linear in parameters, not necessarily in the data. More complex relationships can be quantified using entropy-based metrics such as mutual information. Linear models can also handle interaction terms. We talk about interaction when the model’s output depends on a multiplicative relationship between two or more variables." (Aleksander Molak,Causal Inference and Discovery in Python", 2023)

"Defining an appropriate distance metric from the context ofthe problem is a most important decision. For example, if your variables are all numeric, and on the same scale, then Euclidean distance might be best. If your variables are categorical, you might need to use something like Hamming distance." (Dianne Cook & Ursula Laa, "Interactively Exploring High-Dimensional Data and Models in R", 2026)

"When the ratio of the largest to smallest observation is large you should question whether the data are being analyzed in the right metric" (transformation)." (George E P Box)

🤖Prompt Engineering: Large Language Modeld [LLMs] (Just the Quotes)

"An important capability for our LLM app to become smarter is to maintain a working memory of its interactions with us - otherwise, it will approach each prompt with no knowledge of our previous interactions. For example, it won’t remember details like where we live or what our interests are, which would make it more challenging to develop useful LLM assistants that can use personal information about us to provide more engaging, relevant responses. It also makes it practically more challenging to code a personalized application if we have to explicitly pass context for this personalized information with each interaction, rather than maintaining it 'for free' through LangChain’s memory functionality. It can also allow us to make the LLM specialized for different users by maintaining different memories on different 'threads' that we can visualize and retrieve from LangSmith."(Joseph Babcock & Raghav Bali, "Generative AI with Python and PyTorch" 2nd. Ed., 2025)

"Another problem that can be confusing is that LLMs seldom put out the same thing twice. [...] Traditional databases are straightforward - you ask for something specific, and you get back exactly what was stored. Search engines work similarly, finding existing information. LLMs work differently. They analyze massive amounts of text data to understand statistical patterns in language. The model processes information through multiple layers, each capturing different aspects - from simple word patterns to complex relationships between ideas." (Jeremy C Morgan, "Coding with AI: Examples in Python", 2025)

"As with many other deep learning-based approaches, another major challenge is in interpretability. While knowledge graphs provide a structured and transparent way to store relationships, LLMs operate as a black box, making it difficult to understand how specific outputs are generated. [...] Data alignment is also a key issue, as structured knowledge graphs and unstructured text data must be carefully preprocessed to ensure consistency.  Differences in data formats, ontology mismatches, and information redundancy can create inefficiencies when integrating these two paradigms. Developing robust pipelines that seamlessly connect graph-based insights with LLM-generated text remains an open challenge." (Aldo Marzullo et al, "Graph Machine Learning" 2nd Ed., 2025)

"Despite their impressive capabilities, LLMs are not without limitations. One of the most significant challenges is the problem of hallucination, where an LLM generates factually incorrect or misleading information that appears plausible. This is particularly problematic in domains requiring high factual accuracy, such as healthcare, finance, and legal applications. To mitigate hallucinations and enhance the reliability of LLM outputs,  Retrieval-Augmented Generation (RAG) has emerged as a powerful technique. RAG works by dynamically retrieving relevant information from an external knowledge source (such as a knowledge graph) at inference time, rather than just relying on pre-trained knowledge. This approach ensures that the model has access to up-to-date and accurate data, grounding answers in verified information rather than generating content purely from its internal representations." (Aldo Marzullo et al, "Graph Machine Learning" 2nd Ed., 2025)

"Generative AI for coding and language tools is based on the LLM concept. A large language model is a type of neural network that processes and generates text in a humanlike way. It does this by being trained on a massive dataset of text, which allows it to learn human language patterns, as described previously. It lets LLMs translate, write, and answer questions with text. LLMs can contain natural language, source code, and  more." (Jeremy C Morgan, "Coding with AI: Examples in Python", 2025)

"Generative AI tools for coding are sometimes inaccurate. They can produce results that look good but are wrong. This is common with LLMs. They can write code or chat like a person. And sometimes, they share information that’s just plain wrong. Not just a bit off, but totally backwards or nonsense. And they say it so confidently! We call this 'hallucinating', which is a funny term, but it makes sense." (Jeremy C Morgan, "Coding with AI: Examples in Python", 2025)

"In prompt engineering, we customize the prompts or questions we give the model to get more accurate or insightful responses. The way a prompt is structured has a massive impact on how well a model understands the task at hand and, ultimately, how well it performs. Given LLMs’ versatility, prompt engineering has become an important skill for getting the most out of these models across different domains and tasks. The key is to understand how different prompt structures lead to different model behaviors. There are various strategies - ranging from simple one-shot prompting to more complex techniques like chain-of-thought prompting - that can significantly improve the effectiveness of LLMs." (Abi Aryan, "LLMOps: Managing Large Language Models in Production", 2025)

"It’s essentially a sophisticated prediction system. Instead of looking up stored answers, an LLM calculates probabilities to determine what text should come next. While these predictions are often accurate, they’re still predictions - which is why it’s crucial to verify any code or factual claims the model generates. This probabilistic nature makes LLMs powerful tools for generating text and code but also means they can make mistakes, even when seeming very confident. Understanding this helps set realistic expectations about what these tools can and cannot do reliably."  (Jeremy C Morgan, "Coding with AI: Examples in Python", 2025)

"LLMs are great at generating responses while following instructions but a general empirical observation is a marked improvement in performance when prompts are coupled with a few examples (as opposed to zero-shot scenarios). This is not to say that zero-shot performance is bad but the fact that, in real-life settings, our tasks/requirements are generally a bit more nuanced. For instance, LLMs have an inherent capability to infer sentiment for an input sentence but giving a few examples of how to use that inferred sentiment in responding to customer feedback helps." (Joseph Babcock & Raghav Bali, "Generative AI with Python and PyTorch" 2nd. Ed., 2025)

"LLMs can inadvertently produce toxic content or biased language, leak private information, or be vulnerable to jailbreak prompts. These risks carry serious legal and reputational consequences. To mitigate them, evaluation tools must integrate automated filters and classifiers that flag problematic outputs in real time, as we discussed earlier in the chapter. Metrics such as safety scores, toxicity indices, and bias measurements should be collected alongside model metadata for auditing purposes." (Abi Aryan, "LLMOps: Managing Large Language Models in Production", 2025)

"LLM deployment failures often trace back not to the model itself, but to the prompts it receives. In production environments, prompts are rarely fixed, handcrafted snippets. Instead, they are dynamically generated, assembled from templates, and parameterized based on upstream data sources or evolving user state. This dynamism introduces complexity and variability that can subtly undermine the system’s performance if not carefully managed." (Abi Aryan, "LLMOps: Managing Large Language Models in Production", 2025)

"LLMs excel at understanding context and making associations among words, phrases, and concepts to provide relevant information based on the input query or prompt. While structured knowledge bases rely on humancurated data, LLMs can  automatically extract knowledge from unstructured text. When trained on diverse textual sources, they can process a vast amount of information without explicit human intervention. However, this also introduces a challenge, as the model can learn biased or incorrect information from the training data." (Abi Aryan, "LLMOps: Managing Large Language Models in Production", 2025)

"Prompt engineering is a crucial aspect of working with large language models (LLMs) like OpenAI's GPT, Google's PaLM, and others in the space of AI and machine learning. It involves the art and science of designing inputs (prompts) in a way that maximizes the quality, relevance, and accuracy of the AI-generated output. As the capabilities of AI continue to improve, the task of crafting effective prompts has become an essential skill for anyone leveraging these tools for real-world applications, including natural language understanding, translation, summarization, code generation, and more." (Code Planet, "Python for Large Language Models", 2025)

"RAG is a framework that combines the strengths of traditional information retrieval systems with the generative capabilities of LLMs. In this setup, an LLM is augmented with a retrieval component that fetches relevant information from external data sources, such as knowledge bases or databases, to produce more accurate and contextually relevant responses. This method enhances the LLM’s output by grounding it in authoritative, up-to-date information." (Aldo Marzullo et al, "Graph Machine Learning" 2nd Ed., 2025)

"An alternative future is a more federated open-source AI where commoditised LLM can be run independently by anyone but loosely coordinated between each other if that is mutually advantageous. This requires sophisticated governance of an ecosystem where the rules that maintain the healthy stability of the ecosystem are defined and adhered to. This model allows anyone to put any LLM on a normal computer and plug into the AI. This is the path that already seems most sensible to business leaders and government officials, who recognise that their private data and iterative datasets hold value that can increase productivity and improve the bottom line or assist citizens in their lives." (Alan Watkins & G C Cooke, "Smarter than You Winning in Business with Superintelligent AI", 2026)

"Generative artificial intelligence (GenAI), powered by large language models (LLMs) like Google’s Gemini and OpenAI’s GPT, has transformed how we work and live, revolutionizing business after business. Despite this success, generative AI falls short in domains where specific domain knowledge, high accuracy, and explainability are essential. And it has other significant limitations, including hallucinations and a lack of context and relations. This is where knowledge graphs (KGs) come in, provid-ing contextual information - such as experiences, environmental characteristics, cultural aspects, and social normsneeded to build the 'third wave of AI' for mission-critical applications." (Alessandro Negro et al, "Knowledge Graphs and LLMs in Action", 2026)

"KGs are sophisticated graph structures that represent real-world entities (people, places, diseases, proteins), define meaningful connections between them, and provide context. KGs provide structured, explainable knowledge representation but are challenging to build and query; LLMs offer natural language processing capabilities but suffer from hallucinations, stale information, and a lack of domain-specific grounding. Together, they are a 'killer combination': LLMs can extract entities and relationships from unstructured text to build KGs more efficiently, providing more autonomous and powerful graph querying and analysis. Meanwhile, KGs provide reliable, up-to-date domain knowledge to ground LLM responses and prevent hallucinations." (Alessandro Negro et al, "Knowledge Graphs and LLMs in Action", 2026)

"[...] LLMs raise serious concerns about ethics, bias and fairness, errors in reasoning, hallucinations, and misuse (e.g., misinformation and disinformation). These concerns are exacerbated by modern LLMs being both literal and figurative 'black boxes': Literal black boxes because many advanced AI systems are proprietary and the weights (trained parameters of the models) are not released to the public; and figurative black boxes because even the open-source AI models are so complicated that understanding them and developing safety guardrails has thus far proven extremely difficult." (Mike X Cohen,"50 ML Projects To Understand LLMs", 2026)

"Misinformation is false or misleading information, and its generation by AI is particularly dangerous because of the aura of credibility these systems can project. In a RAG system, misinformation primarily arises from two failure points: Retrieval of inaccurate content from the knowledge base and fabrication or distortion by the large language model (LLM) during generation, even when given good context. The first line of defense is ensuring the integrity of the knowledge base. A RAG system is only as reliable as the documents it has access to. If non-credible, manipulated, or satirical sources are ingested, the system will retrieve and use them as fact. This makes rigorous data curation and source validation the most critical step in combating misinformation. The second line of defense is strengthening the connection between retrieval and generation to prevent the LLM from 'going off script'. The LLM, based on its pre-trained knowledge, might confidently generate an answer that contradicts the provided evidence or adds unsupported details – a phenomenon known as 'hallucination'. To mitigate this, the system must be designed to strictly adhere to the retrieved context." (Saloni Garg et al, "RAG Artificial Intelligence: Retrieval-Augmented Generation in Generative AI", 2026)

"ML is a useful - and under-utilized - framework for studying LLMs. For one thing, LLMs are literally composed of simple ML algorithms (linear weighted averages and nonlinear transformations). Furthermore, using ML techniques like regression, classification, and clustering, can help reveal how concepts like grammar rules are represented inside LLMs. And finally, many people find LLMs to be intimidatingly complicated while finding ML to be much more approachable. Thus, using ML to study LLMs involves using simple tools to understand complicated tools." (Mike X Cohen,"50 ML Projects To Understand LLMs", 2026)

"The foundational premise of RAG is that the most effective way to reduce hallucinations is to provide the LLM with the correct, explicit information needed to answer a query, thereby minimizing its need to rely on fallible parametric knowledge. Therefore, the quality, relevance, and accuracy of the retrieval step are the most significant factors in determining the truthfulness of the final output. Better retrieval is the most powerful antidote to hallucination. If the retriever fails to find the correct information, the generator is essentially left to guess, making hallucinations almost inevitable. The goal is to create a tight, unambiguous link between the user’s question and the evidence in the knowledge base." (Saloni Garg et al, "RAG Artificial Intelligence: Retrieval-Augmented Generation in Generative AI", 2026)

🧭Business Intelligence: Perspectives (Part 30: The Data Science Connection)

Business Intelligence Series

Data Science is a collection of quantitative and qualitative methods, respectively techniques, algorithms, principles, processes and technologies used to analyze, and process amounts of raw and aggregated data to extract information or knowledge it contains. Its theoretical basis is rooted within mathematics, mainly statistics, computer science and domain expertise, though it can include further aspects related to communication, management, sociology, ecology, cybernetics, and probably many other fields, as there’s enough space for experimentation and translation of knowledge from one field to another.  

The aim of Data Science is to extract valuable insights from data to support decision-making, problem-solving, drive innovation and probably it can achieve more in time. Reading in between the lines, Data Science sounds like a superhero that can solve all the problems existing out there, which frankly is too beautiful to be true! In theory everything is possible, when in practice there are many hard limitations! Given any amount of data, the knowledge that can be obtained from it can be limited by many factors - the degree to which the data, processes and models built reflect reality, and there can be many levels of approximation, respectively the degree to which such data can be collected consistently. 

Moreover, even if the theoretical basis seems sound, the data, information or knowledge which is not available can be the important missing link in making any sensible progress toward the goals set in Data Science projects. In some cases, one might be aware of what's missing, though for the data scientist not having the required domain knowledge, this can be a hard limit! This gap can be probably bridged with sensemaking, exploration and experimentation approaches, especially by applying models from other domains, though there are no guarantees ahead!

AI can help in this direction by utilizing its capacity to explore fast ideas or models. However, it's questionable how much the models built with AI can be further used if one can't build mechanistical mental models of the processes reflected in the data. It's like devising an algorithm for winning at lottery small amounts, though investing more money in the algorithm doesn't automatically imply greater wins. Even if occasionally the performance is improved, it's questionable how much it can be leveraged for each utilization. Statistics has its utility when one studies data in aggregation and can predict average behavior. It can’t be used to predict the occurrence of events with a high precision. Think how hard the prediction of earthquakes or extreme weather is by just looking at a pile of data reflecting what’s happening only in a certain zone!

In theory, the more data one has from different geographical areas or organizations, the more robust the models can become. However, no two geographies, respectively no two organizations are alike: business models, the people, the events and other aspects make global models less applicable to local context. Frankly, one has more chances of progress if a model is obtained by having a local scope and then attempting to leverage the respective model for a broader scope. Even then, there can be differences between the behavior or phenomena at micro, respectively at macro level (see the law of physics). 

This doesn’t mean that Data Science or AI related knowledge is useless. The knowledge accumulated by applying various techniques, models and programming languages in problem-solving can be more valuable than the results obtained! Experimentation is a must for organizations to innovate, to extend their knowledge base. It’s also questionable how much of the respective knowledge can be retained and put to good use. In the end, each organization must determine this by itself!

Previous Post <<||>> Next Post

#️⃣Software Engineering: Mea Culpa (Part V: All-Knowing Developers are Back in Demand?)

Software Engineering Series

I’ve been reading many job descriptions lately related to my experience and curiously or not I observed that many organizations look for developers with Microsoft Dynamics experience in the CRM, respectively Finance and Operations (F&O) and Business Central (BC) areas. It’s a good sign that the adoption of Microsoft solutions for CRM and ERP increases, especially when one considers the progress made in the BI and AI areas with the introduction of Microsoft Fabric, which gives Microsoft a considerable boost. Conversely, it seems that the "developers are good for everything" syntagma is back, at least from what one reads in job descriptions. 

Of course, it’s useful to have an inhouse developer who can address all the aspects of an implementation, though that’s a lot to ask considering the different non-programming areas that need to be addressed. It’s true that a developer with experience can handle Requirements, Data and Process Management, respectively Data Migrations and Business Intelligence topics, though if one considers that each of the topics can easily become a full-time job before, during and post-project implementations. I’ve been there and I (hopefully) know that the jobs imply. Even if an experienced programmer can easily handle the different aspects, there will be also times when all the topics combined will be too much for a person!

It's not a novelty that job descriptions are treated like Christmas lists, but it’s difficult to differentiate between essential and nonessential skillset. I read many jobs descriptions lately in which among a huge list of demands, one of the requirements is to program in the F&O framework, sign that D365 programmers are in high demand. I worked for many years as programmer and Software Engineer, respectively in the BI area, where SQL and non-SQL code is needed. Even if I can understand the code in F&O, does it make sense to learn now to program in X++ and the whole framework? 

It's never too late to learn new tricks, respectively another programming language and/or framework. It even helps to provide better solutions in usual areas, though frankly I would invest my time in other areas, and AI-related topics like AI prompting or Data Science seem to be more interesting on the long run, especially when they are already in demand!

There seems to be a tendency for Data Science professionals to do everything, building their own solutions, ignoring the experience accumulated respectively the data models built in BI and Data Analytics areas, as if the topics and data models are unrelated! It’s also true that AI-modeling comes with its own requirements in what concerns data modeling (e.g. translating non-numeric to numeric values), though I believe that common ground can be found!

Similarly, the notebook-based programming seems to replicate logic in each solution, which occasionally makes sense, though personally I wouldn’t recommend it as practice! The other day, I was looking at code developed in Python to mimic the joining of tables, when a view with the same could be easier (re)used, maintained, read and probably more efficient, even if different engines will be used. It will be interesting to see how the mix of spaghetti solutions will evolve over time. There are developers already complaining of the number of objects used in the process by building logic for each layer from the medallion architecture! Even if it makes sense from architectural considerations, it will become a nightmare in time.

One can wonder also about nomenclature used – Data Engineer or Prompt Engineering for the simple manipulation of data between structures in data transformations, respectively for structuring the prompts for AI. I believe that engineering involves more than this, no matter the context! 

Previous Post <<||>> Next Post

🧭Business Intelligence: A Software Engineer's Perspective I (Houston, we have a Problem!)

Business Intelligence Series

One of the critics addressed to the BI/Data Analytics, Data Engineering and even Data Science fields is their resistance to applying Software Engineering (SE) methods in practice. SE can be regarded as the application of sound methods, methodologies, techniques, principles, and practices to obtain high quality economic software in a reproducible manner. At minimum, should be applied SE techniques and practices proven to work, for example the use of best practices, reference technologies, standardized processes for requirements gathering and management, etc. This doesn't mean that one should apply the full extent of SE but consider a minimum that makes sense to adopt.

Unfortunately, the creation of data artifacts (queries, reports, data models, data pipelines, data visualizations, etc.) as process seem to be done after the principle of least action, though least action means here the minimum interaction to push pieces on a board rather than getting the things done. At high level, the process is as follows: get the requirements, build something, present results, get more requirements, do changes, present the results, and the process is repeated ad infinitum.

Given that data artifact's creation finds itself at the intersection of two or more knowledge areas in which knowledge is exchanged in several iterations between the parties involved until a common ground is achieved, this process is totally inefficient from multiple perspectives. First of all, it takes considerably more time than planned to reach a solution, resources being wasted in the process, multiple forms of waste being involved. Secondly, the exchange and retention of knowledge resulting from the process is minimal, mainly on a need by basis. This might look as an efficient approach on the short term, but is inefficient overall.

BI reflects the general issues from SE - most of the issues can be traced back to requirements - if the requirements are incorrect and there's no magic involved in between, then one can't expect for the solution to be correct. The bigger the difference between the initial and final requirements elicited in the process, the more resources are wasted. The more time passes between the start of the development phase and the time a solution is presented to the customer, the longer it takes to build the final solution. Same impact have the time it takes to establish a common ground and other critical factors for success involved in the process.

One can address these issues through better requirements elicitation, rapid prototyping, the use of agile methodologies and similar approaches, though the general feeling is that even if they bring improvements, they don't address the root causes - lack of data literacy skills, lack of knowledge about the business, lack of maturity in planning and executing tasks, the inexistence of well-designed processes and procedures, respectively the lack of an engineering mindset.

These inefficiencies have low impact when building a report occasionally, though they accumulate and tend to create systemic issues in what concerns the overall BI effort. They are addressed locally by experts and in general through a strategic approach like the elaboration of a BI strategy, though organizations seldom pay attention to them. Some organizations consider that they are automatically addressed as part of the data culture, though data culture focuses in general on data literacy and not on the whole set of assumptions mentioned above.

An experienced data professional sees more likely the inefficiencies, tries to address them locally in his interactions with the various stakeholders, he/she can build a business case for addressing them, though it depends on organizations to recognize that they have a problem, respective address the inefficiencies in a strategic and systemic manner!

Previous Post <<||>> Next Post

🧭Business Intelligence: A One-Man Show (Part V: Focus on the Foundation)

Business Intelligence Suite

I tend to agree that one person can't do anymore "everything in the data space", as Christopher Laubenthal put it his article on the topic [1]. He seems to catch the essence of some of the core data roles found in organizations. Summarizing these roles, data architecture is about designing and building a data infrastructure, data engineering is about moving data, database administration is mainly about managing databases, data analysis is about assisting the business with data and reports, information design is about telling stories, while data science can be about studying the impact of various components on the data. 

However, I find his analogy between a college's functional structure and the core data roles as poorly chosen from multiple perspectives, even if both are about building an infrastructure of some type. 

Firstly, the two constructions have different foundations. Data exists in a an organization also without data architects, data engineers or data administrators (DBAs)! It's enough to buy one or more information systems functioning as islands and reporting needs will arise. The need for a data architect might come when the systems need to be integrated or maybe when a data warehouse needs to be build, though many organizations are still in business without such constructs. While for the others, the more complex the integrations, the bigger the need for a Data Architect. Conversely, some systems can be integrated by design and such capabilities might drive their selection.

Data engineering is needed mainly in the context of the cloud, respectively of data lake-based architectures, where data needs to be moved, processed and prepared for consumption. Conversely, architectures like Microsoft Fabric minimize data movement, the focus being on data processing, the successive transformations it needs to suffer in moving from bronze to the gold layer, respectively in creating an organizational semantical data model. The complexity of the data processing is dependent on data' structuredness, quality and other data characteristics. 

As I mentioned before, modern databases, including the ones in the cloud, reduce the need for DBAs to a considerable degree. Unless the volume of work is big enough to consider a DBA role as an in-house resource, organizations will more likely consider involving a service provider and a contingent to cover the needs. 

Having in-house one or more people acting under the Data Analyst role, people who know and understand the business, respectively the data tools used in the process, can go a long way. Moreover, it's helpful to have an evangelist-like resource in house, a person who is able to raise awareness and knowhow, help diffuse knowledge about tools, techniques, data, results, best practices, respectively act as a mentor for the Data Analyst citizens. From my point of view, these are the people who form the data-related backbone (foundation) of an organization and this is the minimum of what an organization should have!

Once this established, one can build data warehouses, data integrations and other support architectures, respectively think about BI and Data strategy, Data Governance, etc. Of course, having a Chief Data Officer and a Data Strategy in place can bring more structure in handling the topics at the various levels - strategical, tactical, respectively operational. In constructions one starts with a blueprint and a data strategy can have the same effect, if one knows how to write it and implement it accordingly. However, the strategy is just a tool, while the data-knowledgeable workers are the foundation on which organizations should build upon!

"Build it and they will come" philosophy can work as well, though without knowledgeable and inquisitive people the philosophy has high chances to fail.

Previous Post <<||>> Next Post

Resources:
[1] Christopher Laubenthal (2024) "Why One Person Can’t Do Everything In Data" (link)

🧭Business Intelligence: A One-Man Show (Part IV: Data Roles between Past and Future)

Business Intelligence Series

Databases nowadays are highly secure, reliable and available to a degree that reduces the involvement of DBAs to a minimum. The more databases and servers are available in an organization, and the older they are, the bigger the need for dedicated resources to manage them. The number of DBAs involved tends to be proportional with the volume of work required by the database infrastructure. However, if the infrastructure is in the cloud, managed by the cloud providers, it's enough to have a person in the middle who manages the communication between cloud provider(s) and the organization. The person doesn't even need to be a DBA, even if some knowledge in the field is usually recommended.

The requirement for a Data Architect comes when there are several systems in place and there're multiple projects to integrate or build around the respective systems. It'a also the question of what drives the respective requirement - is it the knowledge of data architectures, the supervision of changes, and/or the review of technical documents? The requirement is thus driven by the projects in progress and those waiting in the pipeline. Conversely, if all the systems are in the cloud, their integration is standardized or doesn't involve much architectural knowledge, the role becomes obsolete or at least not mandatory. 

The Data Engineer role is a bit more challenging to define because it appeared in the context of cloud-based data architectures. It seems to be related to the data movement via ETL/ELT pipelines and of data processing and preparation for the various needs. Data modeling or data presentation knowledge isn't mandatory even if ideal. The role seems to overlap with the one of a Data Warehouse professional, be it a simple architect or developer. Role's knowhow depends also on the tools involved, because one thing is to build a solution based on a standard SQL Server, and another thing to use dedicated layers and architectures for the various purposes. Engineers' number should be proportional with the number of data entities involved.

Conversely, the existence of solutions that move and process the data as needed, can reduce the volume of work. Moreover, the use of AI-driven tools like Copilot might shift the focus from data to prompt engineering. 

The Data Analyst role is kind of a Cinderella - it can involve upon case everything from requirements elicitation to reports writing and results' interpretation, respectively from data collection and data modeling to data visualization. If you have a special wish related to your data, just add it to the role! Analysts' number should be related to the number of issues existing in organization where the collection and processing of data could make a difference. Conversely, the Data Citizen, even if it's not a role but a desirable state of art, could absorb in theory the Data Analyst role.

The Data Scientist is supposed to reveal the gems of knowledge hidden in the data by using Machine Learning, Statistics and other magical tools. The more data available, the higher the chances of finding something, even if probably statistically insignificant or incorrect. The role makes sense mainly in the context of big data, even if some opportunities might be available at smaller scales. Scientists' number depends on the number of projects focused on the big questions. Again, one talks about the Data Scientist citizen. 

The Information Designer role seems to be more about data visualization and presentation. It makes sense in the organizations that rely heavily on visual content. All the other organizations can rely on the default settings of data visualization tools, independently on whether AI is involved or not. 

Previous Post <<||>> Next Post

Data Science: Back to the Future I (About Beginnings)

Data Science Series

I've attended again, after several years, a webcast on performance improvement in SQL Server with Claudio Silva, “Writing T-SQL code for the engine, not for you”. The session was great and I really enjoyed it! I recommend it to any data(base) professional, even if some of the scenarios presented should be known already.

It's strange to see the same topics from 20-25 years ago reappearing over and over again despite the advancements made in the area of database engines. Each version of SQL Server brought something new in what concerns the performance, though without some good experience and understanding of the basic optimization and troubleshooting techniques there's little overall improvement for the average data professional in terms of writing and tuning queries!

Especially with the boom of Data Science topics, the volume of material on SQL increased considerably and many discover how easy is to write queries, even if the start might be challenging for some. Writing a query is easy indeed, though writing a performant query requires besides the language itself also some knowledge about the database engine and the various techniques used for troubleshooting and optimization. It's not about knowing in advance what the engine will do - the engine will often surprise you - but about knowing what techniques work, in what cases, which are their advantages and disadvantages, respectively on how they might impact the processing.

Making a parable with writing literature, it's not enough to speak a language; one needs more for becoming a writer, and there are so many levels of mastery! However, in database world even if creativity is welcomed, its role is considerable diminished by the constraints existing in the database engine, the problems to be solved, the time and the resources available. More important, one needs to understand some of the rules and know how to use the building blocks to solve problems and build reliable solutions.

The learning process for newbies focuses mainly on the language itself, while the exposure to complexity is kept to a minimum. For some learners the problems start when writing queries based on multiple tables -  what joins to use, in what order, how to structure the queries, what database objects to use for encapsulating the code, etc. Even if there are some guidelines and best practices, the learner must walk the path and experiment alone or in an organized setup.

In university courses the focus is on operators algebras, algorithms, on general database technologies and architectures without much hand on experience. All is too theoretical and abstract, which is acceptable for research purposes,  but not for the contact with the real world out there! Probably some labs offer exposure to real life scenarios, though what to cover first in the few hours scheduled for them?

This was the state of art when I started to learn SQL a quarter century ago, and besides the current tendency of cutting corners, the increased confidence from doing some tests, and the eagerness of shouting one’s shaking knowledge and more or less orthodox ideas on the various social networks, nothing seems to have changed! Something did change – the increased complexity of the problems to solve, and, considering the recent technological advances, one can afford now an AI learn buddy to write some code for us based on the information provided in the prompt.

This opens opportunities for learning and growth. AI can be used in the learning process by providing additional curricula for learners to dive deeper in some topics. Moreover, it can help us in time to address the challenges of the ever-increase complexity of the problems.

Notes: Team Data Science Process (TDSP)

Team Data Science Process (TDSP)
• an agile, iterative data science methodology to deliver predictive analytics solutions and intelligent applications efficiently [1]
• {goal} help customers fully realize the benefits of their analytics program [1]
• {component} data science lifecycle definition
  • {description} a framework to structure the development of data science projects [1]
  • {goal} designed for data science projects that ship as part of intelligent applications that deploy ML & AI models for predictive analytics [1]
  • {benefit} can be used in the context of other DM methodologies as they have common ground [1]
    • e.g. CRISP-DM, KDD
  • {benefit} exploratory data science projects or improvised analytics projects can also benefit from using this process [1]
• {component} standardized project structure
  • {description} a directory structure that includes templates for project documents
    • ⇒makes it easy for team members to find information [1]
    • ⇐templates for the folder structure and required documents are provided in standard locations [1]
    • all code and documents are stored in an agile VCS tracking repository [1]
      • {recommendation} create a separate repository for each project on the VCS for versioning, information security, and collaboration [1]
  • {benefit} organizes the code for the various activities [1]
  • {benefit} allows tracking the progress [1]
  • {benefit} provides checklist with key questions for each project to guarantee process and deliverables’ quality [1]
  • {benefit} enables team collaboration [1]
  • {benefit} allows closer tracking of the code for individual features [1]
  • {benefit} enables teams to obtain better cost estimates [1]
  • {benefit} helps build institutional knowledge across the organization [1]
• {component} recommended infrastructure
  • {description} a set of recommendations for the infrastructure and resources needed for analytics and storage [1]
  • {benefit} addresses cloud and/or on-premises requirements [1]
  • {benefit} enables reproducible analysis [1]
  • {benefit} avoids infrastructure duplication [1]
    • ⇒minimizes inconsistencies and unnecessary infrastructure costs [1]
  • {tools} tools are provided to provision the shared resources, track them, and allow each team member to connect to those resources securely [1]
  • {good practice} create a consistent compute environment [1]
    • ⇐allows team members replicate and validate experiments [1]
• {component} recommended tools and utilities
  • {description} a set of recommendations for the tools and utilities needed for project’s execution [1]
  • {benefit} help lower the barriers and increase the consistency of their adoption [1]
  • {benefit} provides an initial set of tools and scripts to jump-start methodology’s adoption [1]
  • {benefit} helps automate some of the common tasks in the data science lifecycle [1]
    • e.g. data exploration and baseline modeling [1]
  • {benefit} well-defined structure provided for individuals to contribute shared tools and utilities into their team's shared code repository [1]
    • ⇐ resources can then be leveraged by other projects [1]
• {phase} 1: business understanding
  • {goal} define and document the business problem, its objectives, the needed attributes, and the metric(s) used to determine project’s success
  • {goal} identify and document the relevant data sources
  • {step} 1.1: define project’s objectives
    • elicit together with the stakeholders the requirements, define and document the problem and its objectives, respectively the metric(s) used to determine project’s success
      • requires a good understanding of the business processes, data and further characteristics
  • {step} 1.2: identify data sources
    • identify the attributes and the data sources relevant to the problem under study
  • {step} 1.3: define project plan and team*
    • develop a high-level milestone plan and identify the resources needed for executing it
  • {tool} project charter
    • standard template that documents the business problem, the scope of the project, the business objectives and metric(s) used to determine project’s success
• {phase} 2: data acquisition & understanding
  • {goal} prepare the base dataset(s) as needed by the modeling phase into the target repository
  • {goal} build the data ETL/ELT architecture and processes needed for provisioning the basis data
  • {step} 2.1: ingest data
    • make the required data available for the team in the repository where the analytics operations take place
  • {step} 2.2: explore data
    • understand data’s characteristics by leveraging specific tools (visualization, analysis)
    • prepare the data as needed for further processing
  • {step} 2.3: set up pipelines
    • build the pipelines needed for data actualization and qualitative assessment [3]
    • set up a process to score new data or refresh the data regularly [3]
  • {step} 2.4: feasibility analysis*
    • reevaluate the project to determine whether the value expected is sufficient to continue pursuing it
  • {tool} data quality report
    • report that includes data summaries, data mappings, variable ranking, data qualitative assessment(s) and further information [3]
  • {tool} solution architecture
    • diagram and/or textual-based description of the data pipeline(s), technical assumptions and further aspects
  • {tool} data reports
    • document the structure and statistics of the raw data
  • {tool} checkpoint decision
    • decision template document that
      • summarizes the findings of the feasibility analysis step
      • includes a set of choices and recommendations for the next steps
      • serves as basis for the decision on whether to continue or not the project, respectively what the next steps are
• {phase} 3: modeling
  • {goal} create a machine-learning model that addresses the prediction requirements and that's suitable for production
  • {step} 3.1: feature engineering
    • the inclusion, aggregation, and transformation of raw variables to create the features used in the analysis [4]
      • ⇐requires a good understanding of how the features relate to each other and how the ML algorithms use those features [4]
  • {step} 3.2: model selection*
    • choose one or more modeling algorithms that address problem’s characteristics the best
  • {step} 3.3: model training
    • involves the following steps:
      • split the input data into training and test datasets
      • build the models by using the training dataset
      • evaluate the training and the test data set
      • determine the optimal setup and methods
  • {step} 3.4: model evaluation
    • evaluate the performance of the model(s)
  • {step} 3.5: feasibility analysis*
    • evaluate the readiness of the models for use into production, respectively on whether they fulfill project’s objectives
  • {tool} feature sets
    • describe the features developed for the modeling and how they were generated
    • contains pointers to the code used to generate the features
  • {tool} model report
    • a standard, template-based report that provides details on each experiment’s outcomes
    • created for each model tried
  • {tool} checkpoint decision
  • {tool} model performance metrics
    • e.g. ROC curves or MSE
• {phase} 4: deployment
  • {goal} deploy the models and the data pipelines to the environment used for final user acceptance
  • {step} 4.1: operationalize architecture
    • prepare the models and data pipelines for use into production
    • {best practice} expose the models over an open API interface
      • enables models’ consumption from various applications
    • {best practice} build telemetry and monitoring into the models and the data pipelines [5]
      • helps in monitoring and troubleshooting [5]
  • {step} 4.2: deploy solution*
    • deploy the architecture into production
  • {tool} status dashboard
    • displays data on system’s health and key metrics
  • {tool} model report
    • the report in its final form with deployment information
  • {tool} solution architecture
    • the document in its final form
• {phase} 5: customer acceptance
  • {goal} confirm that project’s objectives were fulfilled and get customer’s acceptance
  • {step} 5.1: system validation
    • validate system’s performance and outcomes and confirm that it fulfills customer’s needs
  • {step} 5.2: project signoff*
    • finalize and review documentation
    • handover the solution and afferent documentation to customer
    • evaluate the project against the defined objectives and get customer’ signoff
  • {tool} exit report
  • {tool} technical report
    • contains all the details of the project that are useful for learning about how to operate the system [6]

Acronyms:

Artificial Intelligence (AI)

Cross-Industry Standard Process for Data Mining (CRISP-DM)

Data Mining (DM)

Knowledge Discovery in Databases (KDD)

Team Data Science Process (TDSP) 

Version Control System (VCS)

Visual Studio Team Services (VSTS)

Resources:

[1] Microsoft Azure (2020) What is the Team Data Science Process? [source]

[2] Microsoft Azure (2020) The business understanding stage of the Team Data Science Process lifecycle [source]

[3] Microsoft Azure (2020) Data acquisition and understanding stage of the Team Data Science Process [source]

[4] Microsoft Azure (2020) Modeling stage of the Team Data Science Process lifecycle [source] 

[5] Microsoft Azure (2020) Deployment stage of the Team Data Science Process lifecycle [source]

[6] Microsoft Azure (2020) Customer acceptance stage of the Team Data Science Process lifecycle [source]

🧊Data Warehousing: Architecture (Part III: Data Lakes & other Puddles)

One can consider a data lake as a repository of all of an organization’s data found in raw form, however this constraint might be too harsh as the data found at different levels of processing can be imported as well, for example the results of data mining or other Data Science techniques/methods can be considered as raw data for further processing.

In the initial definition provided by James Dixon, the difference between a data lake and a data mart/warehouse was expressed metaphorically as the transition from bottled water to lakes streamed (artificially) from various sources. It’s contrasted thus the objective-oriented, limited and single-purposed role of the data mart/warehouse in respect to the flow of data in nature that could be tapped and harnessed as desired. These are though metaphors intended to sensitize the buyer. Personally, I like to think of the data lake as an extension of the data infrastructure, in which the data mart or warehouse is integrant part. Imposing further constrains seem to have no benefit.  

Probably the most important characteristic of a data lake is that it makes the data of an organization discoverable and consumable, though from there to insight and other benefits is a long road and requires specific knowledge about the techniques used, as well about organization’s processes and data. Without this data lake-based solutions can lead to erroneous results, same as mixing several ingredients without having knowledge about their usage can lead to cooking experiments aloof from the art of cooking.

A characteristic of data is that they go through continuous change and have different timeliness, respectively degrees of quality in respect to the data quality dimensions implied and sources considered. Data need to reflect the reality at the appropriate level of detail and quality required by the processing application(s), this applying to data warehouses/marts as well data lake-based solutions.

Data found in raw form don’t necessarily represent the true/truth and don’t necessarily acquire a good quality no matter how much they are processed. Solutions need to be resilient in respect to the data they handle through their layers, independently of the data quality and transmission problems. Whether one talks about ETL, data migration or other types of data processing, keeping the data integrity at various levels and layers can be maybe the most important demand upon solutions.

Snapshots as moment-in-time recordings of tables, entities, sets of entities, datasets or whole databases, prove to be often the best mechanisms in keeping data integrity when this aspect is essential to their processing (e.g. data migrations, high-accuracy measurements). Unfortunately, the more systems are involved in the process and the broader span of the solutions over the sources, the more difficult it become to take such snapshots.

A SQL query’s output represents a snapshot of the data, therefore SQL-based solutions are usually appropriate for most of the business scenarios in which the characteristics of data (typically volume, velocity and/or variety) make their processing manageable. However, when the data are extracted by other means integrity is harder to obtain, especially when there’s no timestamp to allow data partitioning on a time scale, the handling of data integrity becoming thus in extremis a programmer’s task. In addition, getting snapshots of the data as they are changed can be a costly and futile task.

Further on, maintaining data integrity can prove to be a matter of design in respect not only to the processing of data, but also in respect to the source applications and the business processes they implement. The mastery of the underlying principles, techniques, patterns and methodologies, helps in the process of designing the right solutions.

Note:
Written as answer to a Medium post on data lakes and batch processing in data warehouses. 

Data Science: Data Strategy (Part II: Generalists vs Specialists in the Field)

Division of labor favorizes the tasks done repeatedly, where knowledge of the broader processes is not needed, where aspects as creativity are needed only at a small scale. Division invaded the IT domains as tools, methodologies and demands increased in complexity, and therefore Data Science and BI/Analytics make no exception from this.

The scale of this development gains sometimes humorous expectations or misbelieves when one hears headhunters asking potential candidates whether they are upfront or backend experts when a good understanding of both aspects is needed for providing adequate results. The development gains tragicomical implications when one is limited in action only to a given area despite the extended expertise, or when a generalist seems to step on the feet of specialists, sometimes from the right entitled reasons. 

Headhunters’ behavior is rooted maybe in the poor understanding of the domain of expertise and implications of the job descriptions. It’s hard to understand how people sustain of having knowledge about a domain just because they heard the words flying around and got some glimpse of the connotations associated with the words. Unfortunately, this is extended to management and further in the business environment, with all the implications deriving from it. 

As Data Science finds itself at the intersection between Artificial Intelligence, Data Mining, Machine Learning, Neurocomputing, Pattern Recognition, Statistics and Data Processing, the center of gravity is hard to determine. One way of dealing with the unknown is requiring candidates to have a few years of trackable experience in the respective fields or in the use of a few tools considered as important in the respective domains. Of course, the usage of tools and techniques is important, though it’s a big difference between using a tool and understanding the how, when, why, where, in which ways and by what means a tool can be used effectively to create value. This can be gained only when one’s exposed to different business scenarios across industries and is a tough thing to demand from a profession found in its baby steps. 

Moreover, being a good data scientist involves having a deep insight into the businesses, being able to understand data and the demands associated with data – the various qualitative and quantitative aspects. Seeing the big picture is important in defining, approaching and solving problems. The more one is exposed to different techniques and business scenarios, with right understanding and some problem-solving skillset one can transpose and solve problems across domains. However, the generalist will find his limitations as soon a certain depth is reached, and the collaboration with a specialist is then required. A good collaboration between generalists and specialists is important in complex projects which overreach the boundaries of one person’s knowledge and skillset. 

Complexity is addressed when one can focus on the important characteristic of the problem, respectively when the models built can reflect the demands. The most important skillset besides the use of technical tools is the ability to model problems and root the respective problems into data, to elaborate theories and check them against reality. 

Complex problems can require specialization in certain fields, though seldom one problem is dependent only on one aspect of the business, as problems occur in overreaching contexts that span sometimes the borders of an organization. In addition, the ability to solve problems seem to be impacted by the diversity of the people involved into the task, sometimes even with backgrounds not directly related to organization’s activity. As in evolution, a team’s diversity is an important factor in achievement and learning, most gain being obtained when knowledge gets shared and harnessed beyond the borders of teams.

Note:
Written as answer to a Medium post on Data Science generalists vs specialists.

🔬Data Science: Evolutionary Algorithm (Definitions)

"An Evolutionary Algorithm (EA) is a general class of fitting or maximization techniques. They all maintain a pool of structures or models that can be mutated and evolve. At every stage in the algorithm, each model is graded and the better models are allowed to reproduce or mutate for the next round. Some techniques allow the successful models to crossbreed. They are all motivated by the biologic process of evolution. Some techniques are asexual (so, there is no crossbreeding between techniques) while others are bisexual, allowing successful models to swap ''genetic' information. The asexual models allow a wide variety of different models to compete, while sexual methods require that the models share a common 'genetic' code." (William J Raynor Jr., "The International Dictionary of Artificial Intelligence", 1999)

"Meta-heuristic optimization approach inspired by natural evolution, which begins with potential solution models, then iteratively applies algorithms to find the fittest models from the set to serve as inputs to the next iteration, ultimately leading to a sub-optimal solution which is close to the optimal one." (Gilles Lebrun et al, "EA Multi-Model Selection for SVM", 2009)

"Evolutionary algorithms are search methods that can be used for solving optimization problems. They mimic working principles from natural evolution by employing a population–based approach, labeling each individual of the population with a fitness and including elements of random, albeit the random is directed through a selection process." (Ivan Zelinka & Hendrik Richter, "Evolutionary Algorithms for Chaos Researchers", Studies in Computational Intelligence Vol. 267, 2010)

"Population-based optimization algorithms in which each member of the population represents a candidate solution. In an iterative process the population members evolve and are then evaluated by a fitness function. Genetic Algorithms and Particle Swarm Optimization are examples of evolutionary algorithms." (Efstathios Kirkos, "Composite Classifiers for Bankruptcy Prediction", 2014)

"A collective term for all variants of (probabilistic) optimization and approximation algorithms that are inspired by Darwinian evolution. Optimal states are approximated by successive improvements based on the variation-selection paradigm. Thereby, the variation operators produce genetic diversity and the selection directs the evolutionary search." (Harish Garg, "A Hybrid GA-GSA Algorithm for Optimizing the Performance of an Industrial System by Utilizing Uncertain Data", 2015)

🔭Data Science: Usefulness (Just the Quotes)

"[…] wrong hypotheses, rightly worked from, have produced more useful results than unguided observation." (Augustus de Morgan, "A Budget of Paradoxes", 1872)

"A false hypothesis, if it serve as a guide for further enquiry, may, at the right stage of science, be as useful as, or more useful than, a truer one for which acceptable evidence is not yet at hand." (William C Dampier, "Science and the Human Mind, Science in the Ancient World", 1912) 

"Rule 1. Original data should be presented in a way that will preserve the evidence in the original data for all the predictions assumed to be useful." (Walter A Shewhart, "Economic Control of Quality of Manufactured Product", 1931)

"Without an adequate understanding of the statistical methods, the investigator in the social sciences may be like the blind man groping in a dark room for a black cat that is not there. The methods of Statistics are useful in an over-widening range of human activities in any field of thought in which numerical data may be had." (Frederick E Croxton & Dudley J Cowden, "Practical Business Statistics", 1937)

"Starting from statistical observations, it is possible to arrive at conclusions which not less reliable or useful than those obtained in any other exact science. It is only necessary to apply a clear and precise concept of probability to such observations. (Richard von Mises, "Probability, Statistics, and Truth", 1939)

"Analogies are useful for analysis in unexplored fields. By means of analogies an unfamiliar system may be compared with one that is better known. The relations and actions are more easily visualized, the mathematics more readily applied, and the analytical solutions more readily obtained in the familiar system." (Harry F Olson, "Dynamical Analogies", 1943)

"The view that machines cannot give rise to surprises is due, I believe, to a fallacy to which philosophers and mathematicians are particularly subject. This is the assumption that as soon as a fact is presented to a mind all consequences of that fact spring into the mind simultaneously with it. It is a very useful assumption under many circumstances, but one too easily forgets that it is false. A natural consequence of doing so is that one then assumes that there is no virtue in the mere working out of consequences from data and general principles." (Alan Turing, "Computing Machinery and Intelligence", Mind Vol. 59, 1950)

"Undoubtedly one of the most elegant, powerful, and useful techniques in modern statistical method is that of the Analysis of Variation and Co-variation by which the total variation in a set of data may be reduced to components associated with possible sources of variability whose relative importance we wish to assess. The precise form which any given analysis will take is intimately connected with the structure of the investigation from which the data are obtained. A simple structure will lead to a simple analysis; a complex structure to a complex analysis." (Michael J Moroney, "Facts from Figures", 1951)

"Extrapolations are useful, particularly in the form of soothsaying called forecasting trends. But in looking at the figures or the charts made from them, it is necessary to remember one thing constantly: The trend to now may be a fact, but the future trend represents no more than an educated guess. Implicit in it is 'everything else being equal' and 'present trends continuing'. And somehow everything else refuses to remain equal." (Darell Huff, "How to Lie with Statistics", 1954)

"The usefulness of the models in constructing a testable theory of the process is severely limited by the quickly increasing number of parameters which must be estimated in order to compare the predictions of the models with empirical results" (Anatol Rapoport, "Prisoner's Dilemma: A study in conflict and cooperation", 1965)

"A model is a useful (and often indispensable) framework on which to organize our knowledge about a phenomenon. […] It must not be overlooked that the quantitative consequences of any model can be no more reliable than the a priori agreement between the assumptions of the model and the known facts about the real phenomenon. When the model is known to diverge significantly from the facts, it is self-deceiving to claim quantitative usefulness for it by appeal to agreement between a prediction of the model and observation." (John R Philip, 1966)

"An average is sometimes called a 'measure of central tendency' because individual values of the variable usually cluster around it. Averages are useful, however, for certain types of data in which there is little or no central tendency." (William A Spirr & Charles P Bonini, "Statistical Analysis for Business Decisions" 3rd Ed., 1967)

"[…] fitting lines to relationships between variables is often a useful and powerful method of summarizing a set of data. Regression analysis fits naturally with the development of causal explanations, simply because the research worker must, at a minimum, know what he or she is seeking to explain." (Edward R Tufte, 'Data Analysis for Politics and Policy", 1974)

"Models, of course, are never true, but fortunately it is only necessary that they be useful. For this it is usually needful only that they not be grossly wrong. I think rather simple modifications of our present models will prove adequate to take account of most realities of the outside world. The difficulties of computation which would have been a barrier in the past need not deter us now." (George E P Box, "Some Problems of Statistics and Everyday Life", Journal of the American Statistical Association, Vol. 74" (365), 1979)

"When a real situation involves chance we have to use probability mathematics to understand it quantitatively. Direct mathematical solutions sometimes exist […] but most real systems are too complicated for direct solutions. In these cases the computer, once taught to generate random numbers, can use simulation to get useful answers to otherwise impossible problems." (Robert Hooke, "How to Tell the Liars from the Statisticians", 1983)

"The fact that [the model] is an approximation does not necessarily detract from its usefulness because models are approximations. All models are wrong, but some are useful." (George Box, 1987)

"The heart of the scientific method is the problem-hypothesis-test process. And, necessarily, the scientific method involves predictions. And predictions, to be useful in scientific methodology, must be subject to test empirically." (Paul Davies, "The Cosmic Blueprint: New Discoveries in Nature's Creative Ability to, Order the Universe", 1988)

"Statistics is a tool. In experimental science you plan and carry out experiments, and then analyse and interpret the results. To do this you use statistical arguments and calculations. Like any other tool - an oscilloscope, for example, or a spectrometer, or even a humble spanner - you can use it delicately or clumsily, skillfully or ineptly. The more you know about it and understand how it works, the better you will be able to use it and the more useful it will be." (Roger Barlow, "Statistics: A Guide to the Use of Statistical Methods in the Physical Sciences", 1989)

"Science is (or should be) a precise art. Precise, because data may be taken or theories formulated with a certain amount of accuracy; an art, because putting the information into the most useful form for investigation or for presentation requires a certain amount of creativity and insight." (Patricia H Reiff, "The Use and Misuse of Statistics in Space Physics", Journal of Geomagnetism and Geoelectricity 42, 1990)

"A useful description relates the systematic variation to one or more factors; if the residuals dwarf the effects for a factor, we may not be able to relate variation in the data to changes in the factor. Furthermore, changes in the factor may bring no important change in the response. Such comparisons of residuals and effects require a measure of the variation of overlays relative to each other." (Christopher H Schrnid, "Value Splitting: Taking the Data Apart", 1991)

"In constructing a model, we always attempt to maximize its usefulness. This aim is closely connected with the relationship among three key characteristics of every systems model: complexity, credibility, and uncertainty. This relationship is not as yet fully understood. We only know that uncertainty" (predictive, prescriptive, etc.) has a pivotal role in any efforts to maximize the usefulness of systems models. Although usually" (but not always) undesirable when considered alone, uncertainty becomes very valuable when considered in connection to the other characteristics of systems models: in general, allowing more uncertainty tends to reduce complexity and increase credibility of the resulting model. Our challenge in systems modelling is to develop methods by which an optimal level of allowable uncertainty can be estimated for each modelling problem." (George J Klir & Bo Yuan, "Fuzzy Sets and Fuzzy Logic: Theory and Applications", 1995)

"The point is that scientific descriptions of phenomena in all of these cases do not fully capture reality they are models. This is not a shortcoming but a strength of science much of the scientist's art lies in figuring out what to include and what to exclude in a model, and this ability allows science to make useful predictions without getting bogged down by intractable details." (Philip Ball, "The Self-Made Tapestry: Pattern Formation in Nature", 1998)

"We should push for de-emphasizing some topics, such as statistical significance tests - an unfortunate carry-over from the traditional elementary statistics course. We would suggest a greater focus on confidence intervals - these achieve the aim of formal hypothesis testing, often provide additional useful information, and are not as easily misinterpreted." (Gerry Hahn et al, "The Impact of Six Sigma Improvement: A Glimpse Into the Future of Statistics", The American Statistician, 1999)

"The point of a model is to get useful information about the relation between the response and predictor variables. Interpretability is a way of getting information. But a model does not have to be simple to provide reliable information about the relation between predictor and response variables; neither does it have to be a data model. The goal is not interpretability, but accurate information." (Leo Breiman, "Statistical Modeling: The Two Cultures, Statistical Science" Vol. 16(3), 2001)

"Models can be viewed and used at three levels. The first is a model that fits the data. A test of goodness-of-fit operates at this level. This level is the least useful but is frequently the one at which statisticians and researchers stop. For example, a test of a linear model is judged good when a quadratic term is not significant. A second level of usefulness is that the model predicts future observations. Such a model has been called a forecast model. This level is often required in screening studies or studies predicting outcomes such as growth rate. A third level is that a model reveals unexpected features of the situation being described, a structural model, [...] However, it does not explain the data." (Gerald van Belle, "Statistical Rules of Thumb", 2002)

"[…] studying methods for parametric models is useful for two reasons. First, there are some cases where background knowledge suggests that a parametric model provides a reasonable approximation. […] Second, the inferential concepts for parametric models provide background for understanding certain nonparametric methods." (Larry A Wasserman, "All of Statistics: A concise course in statistical inference", 2004)

"A model is a simplification or approximation of reality and hence will not reflect all of reality. […] Box noted that ‘all models are wrong, but some are useful’. While a model can never be ‘truth’, a model might be ranked from very useful, to useful, to somewhat useful to, finally, essentially useless.”" (Kenneth P Burnham & David R Anderson, 'Model Selection and Multimodel Inference: A Practical Information-Theoretic Approach” 2nd Ed., 2005)

"First, we affirm that all models are wrong, some of them are useful. Since a model is an abstraction of reality, and that too only from a particular perspective, they are fundamentally wrong because they are not reality. That gives no license to models that are wrongly built - after all, two wrongs don’t make a right. So usefulness, or purpose, is what determines a model’s role, given that it is correctly formed. Models therefore have teleological value even though they are ontologically erroneous." (John Boardman & Brian Sauser, "Systems Thinking: Coping with 21st Century Problems", 2008)

"We face danger whenever information growth outpaces our understanding of how to process it. The last forty years of human history imply that it can still take a long time to translate information into useful knowledge, and that if we are not careful, we may take a step back in the meantime." (Nate Silver, "The Signal and the Noise", 2012)

"Complexity scientists concluded that there are just too many factors - both concordant and contrarian - to understand. And with so many potential gaps in information, almost nobody can see the whole picture. Complex systems have severe limits, not only to predictability but also to measurability. Some complexity theorists argue that modelling, while useful for thinking and for studying the complexities of the world, is a particularly poor tool for predicting what will happen." (Lawrence K Samuels, "Defense of Chaos: The Chaology of Politics, Economics and Human Action", 2013)

"In general, when building statistical models, we must not forget that the aim is to understand something about the real world. Or predict, choose an action, make a decision, summarize evidence, and so on, but always about the real world, not an abstract mathematical world: our models are not the reality - a point well made by George Box in his oft-cited remark that “all models are wrong, but some are useful." (David Hand, "Wonderful examples, but let's not close our eyes", Statistical Science 29, 2014)

"To find signals in data, we must learn to reduce the noise - not just the noise that resides in the data, but also the noise that resides in us. It is nearly impossible for noisy minds to perceive anything but noise in data. […] Signals always point to something. In this sense, a signal is not a thing but a relationship. Data becomes useful knowledge of something that matters when it builds a bridge between a question and an answer. This connection is the signal." (Stephen Few, "Signal: Understanding What Matters in a World of Noise", 2015)

"Data visualization is marketed today as the miracle cure that will open the doors to success, whatever its shape. We have enough experience to realize that in reality it’s not always easy to distinguish between real usefulness and zealous marketing. After the initial excitement over the prospects of data visualization comes disillusionment, and after that the possibility of a balanced assessment. The key is to get to this point quickly, without disappointments and at a lower cost." (Jorge Camões, "Data at Work: Best practices for creating effective charts and information graphics in Microsoft Excel", 2016)

"Effects without an understanding of the causes behind them, on the other hand, are just bunches of data points floating in the ether, offering nothing useful by themselves. Big Data is information, equivalent to the patterns of light that fall onto the eye. Big Data is like the history of stimuli that our eyes have responded to. And as we discussed earlier, stimuli are themselves meaningless because they could mean anything. The same is true for Big Data, unless something transformative is brought to all those data sets… understanding." (Beau Lotto, "Deviate: The Science of Seeing Differently", 2017)

"The goal of data science is to improve decision making by basing decisions on insights extracted from large data sets. As a field of activity, data science encompasses a set of principles, problem definitions, algorithms, and processes for extracting nonobvious and useful patterns from large data sets. It is closely related to the fields of data mining and machine learning, but it is broader in scope." (John D Kelleher & Brendan Tierney, "Data Science", 2018)

"In machine learning, our data has biases as well as useful information for our task. The more exactly our machine learning model fits the data, the more it reflects these biases. This means that the predictions may be based on spurious relationships that incidentally occur in the training data." (Alex Thomas, "Natural Language Processing with Spark NLP", 2020)

🔭Data Science: Statistical Tests (Just the Quotes)

"The practical power of a statistical test is the product of its statistical power and the probability of use." (John W Tukey, A Quick, "Compact, Two Sample Test to Duckworth’s Specifications", 1959)

"Statistical significance testing has involved more fantasy than fact. The emphasis on statistical significance over scientific significance in educational research represents a corrupt form of the scientific method. Educational research would be better off if it stopped testing its results for statistical significance. (Ronald P. Carver,The case against statistical testing", Harvard Educational Review 48, 1978)

"Statistical significance testing can involve a tautological logic in which tired researchers, having collected data on hundreds of subjects, then conduct a statistical test to evaluate whether there were a lot of subjects, which the researchers already know, because they collected the data and know they are tired. This tautology has created considerable damage as regards the cumulation of knowledge." (Bruce Thompson,Two and One-Half Decades of Leadership in Measurement and Evaluation", Journal of Counseling & Development 70" (3), 1992)

"The practical definitions of randomness - a sequence is random by virtue of how many and which statistical tests it satisfies and a sequence is random by virtue of the length of the algorithm necessary to describe it [...]." (Deborah J Bennett, "Randomness", 1998)

"First, if you already know that the population from which your sample has been taken is normally distributed (perhaps you have data for a variable that has been studied before), you can assume the distribution of sample means from this population will also be normally distributed. Second, the central limit theorem […] states that the distribution of the means of samples of about 25 or more taken from any population will be approximately normal, provided the population is not grossly non-normal (e.g. a population that is bimodal). Therefore, provided your sample size is sufficiently large you can usually do a parametric test. Finally, you can examine your sample. Although there are statistical tests for normality, many statisticians have cautioned that these tests often indicate the sample is significantly non normal even when a t-test will still give reliable results." (Steve McKillup, "Statistics Explained: An Introductory Guide for Life Scientists", 2005)

"Statistical tests are just a way of working out the probability of obtaining the observed, or an even more extreme, difference among samples (or between an observed and expected value) if a specific hypothesis (usually the null of no difference) is true. Once the probability is known, the experimenter can make a decision about the difference, using criteria that are uniformly used and understood." (Steve McKillup, "Statistics Explained: An Introductory Guide for Life Scientists", 2005)

"What is so unconventional about the statistical way of thinking? First, statisticians do not care much for the popular concept of the statistical average; instead, they fixate on any deviation from the average. They worry about how large these variations are, how frequently they occur, and why they exist. [...] Second, variability does not need to be explained by reasonable causes, despite our natural desire for a rational explanation of everything; statisticians are frequently just as happy to pore over patterns of correlation. [...] Third, statisticians are constantly looking out for missed nuances: a statistical average for all groups may well hide vital differences that exist between these groups. Ignoring group differences when they are present frequently portends inequitable treatment. [...] Fourth, decisions based on statistics can be calibrated to strike a balance between two types of errors. Predictably, decision makers have an incentive to focus exclusively on minimizing any mistake that could bring about public humiliation, but statisticians point out that because of this bias, their decisions will aggravate other errors, which are unnoticed but serious. [...] Finally, statisticians follow a specific protocol known as statistical testing when deciding whether the evidence fits the crime, so to speak. Unlike some of us, they don’t believe in miracles. In other words, if the most unusual coincidence must be contrived to explain the inexplicable, they prefer leaving the crime unsolved." (Kaiser Fung, "Numbers Rule the World", 2010) 

"Another way to secure statistical significance is to use the data to discover a theory. Statistical tests assume that the researcher starts with a theory, collects data to test the theory, and reports the results - whether statistically significant or not. Many people work in the other direction, scrutinizing the data until they find a pattern and then making up a theory that fits the pattern." (Gary Smith,Standard Deviations", 2014)

"Null hypothesis is something we attempt to find evidence against in the hypothesis tests. Null hypothesis is usually an initial claim that researchers make on the basis of previous knowledge or experience. Alternative hypothesis has a population parameter value different from that of null hypothesis. Alternative hypothesis is something you hope to come out to be true. Statistical tests are performed to decide which of these holds true in a hypothesis test. If the experiment goes in favor of the null hypothesis then we say the experiment has failed in rejecting the null hypothesis." (Danish Haroon,Python Machine Learning Case Studies", 2017)