5 Reasons You Need Statistics for Machine Learning


Machine learning is a branch of data science that is getting all the attention right now. Some people try to tap into this by adding proficiency in the field to their resume. Unfortunately, not many of them have the right foundation, so they often create models that fail.

You need Statistics for machine learning because with a decent understanding of statistical methods you can convert raw observations into information that is easy to understand, digest, and share. This will allow you to create machine learning models that will consistently deliver results.

In this article, you’ll see the various reasons you can’t ignore statistics if you are thinking about becoming a robust machine learning professional. Watch out for recommended books that can help when you start learning statistics to become better grounded in machine learning.

Important Sidenote: We interviewed 100+ data science professionals (data scientists, hiring managers, recruiters – you name it) and identified 6 proven steps to follow for becoming a data scientist. Read my article: ‘6 Proven Steps To Becoming a Data Scientist [Complete Guide] for in-depth findings and recommendations! – This is perhaps the most comprehensive article on the subject you will find on the internet!

What Is Statistics?

Statistics is an arm of applied mathematics which encapsulates the collection, organization, analysis, interpretation, and presentation of data. It helps you to glean answers to important questions posed by a data cache.

In the wider scope of data science, descriptive statistical methods convert raw observations into actionable insights. In contrast, inferential statistical methods can work on everything from small samples of data to entire domains.

Do You Need to Know Statistics for Machine Learning?

Yes, you need statistics for machine learning. Both fields of study are highly intertwined, to the point that some statisticians refer to machine learning as statistical learning or applied statistics—instead of the name that is designed to sound a bit more computer-centric.

When getting started with machine learning, the bulk of the texts assume that you already have some statistics foundation, highlighting how it’s hard to have a sound foundation in machine learning without it. Some examples of machine learning books that explicitly require knowledge of statistics include the following:

These are just some examples showing that you need some basic understanding of statistics to properly understand machine learning. Almost anyone can apply an algorithm lifted off different sources to a dataset and claim proficiency in machine learning.

However, without adequate knowledge of statistics, you’ll find out that you can’t interpret logistic regression results. You’ll also see a poor performance from your models because you’ve failed to normalize predictors, and you’re likely using the incorrect splitting criterion with your tree-based models. You need a proper background in statistics to avoid these problems.

Most Machine Learning Books Require Knowledge of It

We’ve just touched on three books where the authors explicitly mention that they expect you to have some statistics background to grasp the concepts discussed. These are just a few examples. Most books on the subject will have the same approach. Even when they don’t spell it out, you’re almost certain to find many concepts that will be hard to grasp without adequate knowledge of statistics.

You Need Statistics to Convert Data to Information

Raw observations are just data. They are not pieces of information or knowledge. With every dataset, there are a few questions that have to be answered: What does the data look like? Are there any limits on the observation? What observation is most common? 

Away from raw data, you may need to design an experiment that will help you to collect observations. The result of the experiment will raise more questions like the difference in the outcome of the two experiments and whether these differences are noise in the data or real. You’ll also need to know what variables in the experiment are most relevant.

By answering these questions, you can turn the raw observation into usable information. The results generated will be vital to the project. It will also matter to your stakeholders because the information generated will ensure better decision making overall.

So, to understand the data used in training a machine learning model and properly interpret the results, you’ll need statistics. Every step in a typical predictive modeling project will involve some use of a statistical method.

Statistics Concepts and Terminologies Are Used in Machine Learning

Below are some of the common statistics concepts and terminologies you’ll come across in machine learning:

Basic Terminologies

Some of the common statistics terminologies you should expect to see in machine learning include:

  • Statistical parameter: This is the quantity that culminates in probability distributions such as mode, mean, and median.
  • Variable: This is an item of data that can be measured, usually a number.
  • Population: This refers to the source of a dataset or where it’s derived from.
  • Sample: This is the subset of the population.

Statistical Analysis

This refers to the practice of exploring a large collection of datasets to find hidden patterns and trends. You’ll find this type of analysis in use when there are decisions to be modeled. There are two major types of statistical analysis:

  • Qualitative analysis: It involves the why and how of a decision. The data used can come in texts, images, sounds, and more.
  • Quantitative analysis: This type of analysis solves the what, where, and when. It involves gathering and interpreting data using graphs and charts to uncover any underlying trends.

Measures of Central Tendency

This is a single value that aims to describe a dataset by pinpointing the central position within the group. It may also be referred to in some textbooks as a “measure of central location” or categorized under summary statistics. The values are referred to in three ways:

  • Mode: This is the value that appears the most in any dataset.
  • Mean: This is a sum of all the values present in a dataset, divided by the number of values in the set.
  • Median: This is the middle value in the dataset. It is one of the most useful values in the analysis because it is not affected by the skewness of the data or influenced by outliers.

Skewness

Skewness refers to a curve that is tilted towards the left or to the right. It shows the data distribution and helps the analyst see if the data is more intensive on any specific side. 

Skewness can be positive or negative. With positive skewness, the tail of the curve will appear skewed to the right. You’ll also have the relationship between the mean, median, and mode as mean>median<mode. The relationship will appear as mean<median<mode with negative skewness, with the tail of the curve tilted to the left.

Probability

This is the base of statistics and one of the concepts you’ll use a lot in machine learning. It’s defined as the likelihood of a specific outcome and its overall importance. It’s almost impossible to work in machine learning or data science in general without showing a good understanding of probability. It’s one of the pillars of predictive analytics.

Statistical Hypothesis

This is a statement about the nature of a population. It’s divided into null or alternative hypotheses. With a null hypothesis, there is no definitive difference between the described population. On the other hand, an alternative hypothesis suggests there’s a notable difference. 

Linear Regression

Linear regression is the cornerstone of statistics. It is used to predict a variable’s value by incorporating the value of other variables within the dataset. Linear regression is divided into simple and multiple. In multiple regression, more than one independent variable is used to predict the value of the dependent variable. This ensures better accuracy overall.

Classification

This term refers to data mining methods where the available datasets are categorized to derive accurate predictions and analysis overall. Also known as a “Decision Tree,” classification can be done in two methods: Logistic Regression and Discriminant Analysis.

Resampling Methods

Here, samples are drawn from the original dataset to generate a unique sampling distribution that mirrors the main data set. Resampling comes into play when you’re dealing with a dataset that can’t be analyzed whole—as you’ll find with most big data projects. Resampling methods break down the task at hand, but the estimate obtained remains unbiased.

You Need Statistics to Deal With Domain-Focused Problems

Any data scientist in a product-based environment will be expected to support crucial decisions with his model results and analysis. This requires a strong understanding of the domain. For domains that rely heavily on computation, a strong statistics and mathematics foundation is crucial.

For instance, if you’re a data scientist employed as a quant in a hedge fund to develop models to price derivatives and securities, you need to understand how log returns, calculus, and normal distribution can contribute to the development of the model.

Research-facing multi-billion dollar domains like drug discovery will also demand heavy use of conventional statistical concepts. You need to know concepts such as skewness, standard deviation, mean, bootstrapping, sampling, kurtosis, and more.

Statistics Is a Requirement in Job Descriptions

When you read through job descriptions for data science positions that will involve machine learning, you’ll find a demand for expertise in statistical data analysis. 

You need to know statistics well enough to tackle some of the typical problems you’ll be presented with during an interview. Things like knowing how a linear regression model can be optimized and understanding how a decision tree calculates impurity at each node can only come with proper statistics and math background.

These are some of the top reasons why you need statistics for machine learning. You can get started with statistics for machine learning by taking courses online. If you’d like to read some books instead, we’ll cover a few quality options below.

Best Books on Statistics for Machine Learning

Some of the best books on statistics for machine learning you can get today include:

Introduction to Statistical Learning

This book is one of the most recommended for practical data science. It does a great job of linking statistics concepts and machine learning. You’ll learn a lot about supervised and unsupervised machine learning algorithms. 

The book’s practical aspect is demonstrated using R, so you’ll enjoy some advantage if you’re an R user. The book doesn’t only talk about the theoretical side of things. It emphasizes the use of machine learning algorithms in real-life applications.

Elements of Statistical Learning

Written by Stanford University professors, this book is all about exposing you to higher-level algorithms, including Kernel methods, Bagging and Boosting, Neural Networks, and more. Again, the algorithms covered have been implemented in R.

Think Stats

This book focuses on performing statistical analysis in Python. So, you should have some basic knowledge of the Python language before choosing to go with this book. It does an excellent job of helping the reader to understand how statistics influences real life, with the aid of popular and relatable case studies. The book also has some chapters dedicated to topics that are more math than statistics-leaning, such as Bayesian estimation. 

From Algorithms to Z Scores

Do you fully understand the importance of statistics in programming? University of California’s Professor Norm Matloff uses statistical measures in R and some probabilistic concepts to help you understand statistics in programming. You’ll learn how to deal with probabilistic models and how to choose the best of the lot for final evaluation. It’s another great book for your library if you’re an R user.

Introduction to Bayesian Statistics

This book is highly recommended for newcomers to data science. It’s a good book to read if you find mathematics boring, thanks to the conversational style. It is a good introductory resource on statistics. The first part of the book explores scientific methods of data gathering, while the last part covers Bayesian statistics.

Discovering Statistics Using R

This book is another excellent work that will be of great help to newcomers in data science. The authors go into great detail in all the topics covered. Like other books we have looked at thus far, statistical concepts are explained in R. The stimulating practice examples make the book a more effective resource for grasping statistics.

Naked Statistics: Stripping the Dread From the Data

If you’ve always found it difficult to stick with statistics classes due to the technical details, you’ll love this book. The author gets rid of all the technical details and emphasizes the main intuition, which drives statistical analysis.

The book sheds light on concepts like regression analysis, correlation, and inference. It also talks about how easy it is for carelessness to cause misrepresentation or manipulation of data. The examples in the book show the creative ways researchers are using data to deal with various problems.

All of Statistics: A Concise Course in Statistical Inference

This book is important because it gives proper attention to topics typically relegated to the background or recommended as follow-up options in other books. The topics include classification, bootstrapping, and nonparametric curve estimation. The author expects you to know linear algebra and calculus, but you can still enjoy the book without previous knowledge of statistics and probability.

Practical Statistics for Data Scientists: 50 Essential Concepts

This is a practical guide that teaches how to transfer different statistical methods to data science. You’ll get advice on what is necessary and what you can ignore and learn how to avoid misusing concepts. The book requires some exposure to statistics and knowledge of R programming knowledge. It is the perfect choice to go with if you’re looking for an accessible and readable resource that will provide you with a truly statistical perspective.

Should You Get a Statistics Degree for Machine Learning?

As you’ve seen above, most of the statistics concepts you’ll need to learn for machine learning are topics you’ll learn from books and online courses with a little discipline and application. You don’t need to spend money on a degree for these concepts.

However, the situation will vary from one person to another. If you have some time and can find some affordable courses in a reputable institute around you, getting a statistics degree isn’t a bad idea. The extra degree will definitely add more gloss to your existing qualifications.

Author’s Recommendations: Top Data Science Resources To Consider

Before concluding this article, I wanted to share few top data science resources that I have personally vetted for you. I am confident that you can greatly benefit in your data science journey by considering one or more of these resources.

  • DataCamp: If you are a beginner focused towards building the foundational skills in data science, there is no better platform than DataCamp. Under one membership umbrella, DataCamp gives you access to 335+ data science courses. There is absolutely no other platform that comes anywhere close to this. Hence, if building foundational data science skills is your goal: Click Here to Sign Up For DataCamp Today!
  • MITx MicroMasters Program in Data Science: If you are at a more advanced stage in your data science journey and looking to take your skills to the next level, there is no Non-Degree program better than MIT MicroMasters. Click Here To Enroll Into The MIT MicroMasters Program Today! (To learn more: Check out my full review of the MIT MicroMasters program here)
  • Roadmap To Becoming a Data Scientist: If you have decided to become a data science professional but not fully sure how to get started: read my article – 6 Proven Ways To Becoming a Data Scientist. In this article, I share my findings from interviewing 100+ data science professionals at top companies (including – Google, Meta, Amazon, etc.) and give you a full roadmap to becoming a data scientist.

Conclusion

Statistics is an integral part of machine learning and data science in general. Unfortunately, not many scientists have formal statistics training. The group with a proper foundation in statistics tends to shine through amid competition, making them the first choice of recruiters.

With a good foundation in statistics, you’ll be able to deliver more robust machine-learning solutions, becoming a major resource person in your domain. We’ve covered some books that can help you get started in statistics for machine learning and data science but do not hesitate to enroll in an online class or for a degree if necessary.

BEFORE YOU GO: Don’t forget to check out my latest article – 6 Proven Steps To Becoming a Data Scientist [Complete Guide]. We interviewed 100+ data science professionals (data scientists, hiring managers, recruiters – you name it) and created this comprehensive guide to help you land that perfect data science job.

  1. Meserole, C. (2019, October 25). What is machine learning? Brookings. https://www.brookings.edu/research/what-is-machine-learning/
  2. Pearson mode skewness: Definition and formulas. (2020, September 16). Statistics How To. https://www.statisticshowto.com/pearson-mode-skewness/
  3. Statistics books for machine learning. (2020, August 14). Machine Learning Mastery. https://machinelearningmastery.com/statistics-books-for-machine-learning/
  4. Statistics for data science: Basic concepts. (n.d.). Hackr.io. https://hackr.io/blog/statistics-for-data-science
  5. Tan, J. (2020, July 9). Statistics and Machine Learning — When to use what? Medium. https://medium.com/swlh/statistics-and-machine-learning-when-to-use-what-bf687bf8eb3
  6. What is machine learning? (2020, June 26). I School Online – UC Berkeley School of Information. https://ischoolonline.berkeley.edu/blog/what-is-machine-learning/
  7. What is statistics? (n.d.). https://www.stat.uci.edu/what-is-statistics/
  8. Why statistics is important for mastering AI/ML skills. (2019, October 19). Analytics India Magazine. https://analyticsindiamag.com/why-statistics-is-importantfor-mastering-ai-ml-skills/

Affiliate Disclosure: We participate in several affiliate programs and may be compensated if you make a purchase using our referral link, at no additional cost to you. You can, however, trust the integrity of our recommendation. Affiliate programs exist even for products that we are not recommending. We only choose to recommend you the products that we actually believe in.

Daisy

Daisy is the founder of DataScienceNerd.com. Passionate for the field of Data Science, she shares her learnings and experiences in this domain, with the hope to help other Data Science enthusiasts in their path down this incredible discipline.

Recent Posts