Do you need math/stat to be a successful data scientist?
If you are starting or switching your career into data science, you often stumble upon this question “do I need math to be a successful data scientist?” More so, when you are not so good at math and still want to venture into data science field. When I had begun my journey as a data scientist, I often encountered this question on several blogs and from several new enthusiasts trying to set their foot on to this famously acclaimed “sexiest job of 21st century.” Many of you might even have come across some form of this famous 3 circles (sometimes expanded to 4 or more) diagram where 1 circle is almost always math and statistics as seen below.
Most of the statistician turned data scientists would swear to god that it is a quintessential skill to be a successful data scientist (not those data scientists who think that data science is all about fancy models and accuracy scores). But even before digging deeper into how much math is needed, an absolute clarity on what kind of data scientist you want to be is a must. In my experience, this is faaaaaaaaaaaaaar more important question to ask than how much math and if math is needed.
In my opinion, there are at least 3 major types of data scientists:
1. That expert data scientist who created Venn diagrams like above: A data science practitioner who can build his/her own algorithms to solve the complex challenges. This kind of data scientist reaches out to a pen and paper to write the logical structure/pseudocode for their algorithm first and then codify their logic. For them the programming language in which they code is not important, the logic and the equation is!
2. A citizen data scientist who understands concepts of data science and can apply and to some extent tweak the algorithms written by first kind of data scientists. These data scientists are the ones who are familiar with several packages of a particular programming language and have mastered application of them. Their best friends are google and stackoverflow. Their logic might not be the best but they get you the solution. It is easy for these kind to fall into the accuracy prey and other evaluation metrics
3. A data science translator whose skills are little similar to citizen data scientists, but these kind are the ones who have other areas of expertise like strong domain knowledge, business acumen, deep understanding of the business problem and the stakeholders’ needs. For them, data science/technology is just a medium to solve the business problem at hand. They focus more on interpretability and explainability of the models and its ability to solve the question at hand rather than accuracy, math,code, algorithm or model.
Once you are clear on what kind you are or want to be, it is easy to find out how much math you need. If you are speaking to only one kind, you might get a biased opinion on how much math is needed. The expert data scientists often take pride in their ability to build their own equations and algorithms and see Math as center of everything in data science. They are generally PhD’s and only value you as a data scientist if you have sound math/stat knowledge and (usually) a PhD degree. The second kind will tell you that math is hype, just do the basics and learn essentials like linear algebra, calculus, probability etc.. The third type will tell you just learn what you can to get started and understand how they work logically and focus more on problem solving.
If you have decided to be either 2nd or 3rd kind, few resources I found very intuitive and helpful are:
3 Blue 1 Brown
Khan Academy
Mathematics for Machine Learning
Its now time to draw your own Venn diagram based on your kind. I am the third kind and have been so far successful in bridging the gap between tech and business. This is how my Venn looks like :)
My key success factor was to be extremely good at 1–2 skills and be conversant in many others (more diversification the better it is as you can speak to several audience)
