Related Posts
Bain & Company My 13 YO is taking Career and Tech. She is dead set on diving for Michigan. What are her chances of getting hired from MBB with a degree from UM? Or would she need her MBA first? Is a Michigan degree good enough to get you into a top MBA program?
McKinsey & Company Boston Consulting Group Bain & Company University of Michigan
More Posts
Why are relationships so difficult!!!
Stuck in same company for 7 years without growth, working in production support without skilling in latest technologies. Started learning React and want to shift on React frontend path. Again, stuck at resume building i dont know how to incorporate react into the projects i have worked. Can anyone send me (n4bryqzp@duck.com) React resume for reference?
Additional Posts in Data & Analytics Consultants
What’s everyone’s take on Alteryx?
New to Fishbowl?
unlock all discussions on Fishbowl.





Most of data science uses these handful of models. Know them and the math behind it well
My take is to get really knowledgeable about linear regression and it’s various extensions (GLM, mixed effect models, GAM, etc). Learning what assumptions need to be in place will help you learn to think about the data and how it is generated. Once you have that knowledge switching to a model like random forest or GBM or other ML based models is basically a drop in replacement.
This, this, this. Honestly OP, I’d recommend an intro to Econometrics course. Although you may be using STATA or R instead of Python and won’t touch SQL, you’ll get an extremely in depth overview of regressions, why they matter and what they do mathematically, and how to interpret and explain the results. Econometrics truly changed how I see the world and data
Bowl Leader
Anything probability is fair game, especially if the role emphasizes A/B testing (show your work level understanding). As distributions go, if you can explain what a marginal distribution is and define the power of a test, you should be golden. Be able to derive matrix OLS by hand, and to code up Naïve Bayes. Know how to interpret regression and logistic regression values.
Basic ML concepts like D1 said, though you don't have to go too deep on the math: I know almost no one who can explain what gradient boosting does at a non-superficial level, and random forests are a stretch...
Pay special attention to evaluation metrics (including non-performance ones like cross-entropy) and related curves. Be able to explain cross validation, bonus points for knowing the limitations.
Other misc stuff... Gradient descent by hand and by code. Solve a system of equations by hand, be able to do basic differentiation. Basic theory of time series models (ARIMA and friends) if appropriate to the role. Minimal graph math if appropriate to the role.
Yes, it's a lot. Good news though: not everything is asked in every interview. Prep everything, but don't sweat it if you miss on something. One, no one expects you to know literally everything. Two, you'll eventually make it through by stumbling on an interview that happens to ask the "right" questions.
More good news: stuff that you can almost certainly skip. Questions on explicit linear algebra, calc past derivatives, and stats derivations beyond OLS are highly unlikely. Harder stats concepts like what to do with beta/gamma distributions, what are copulas, or even any Bayesian stuff beyond Bayes theorem may come up in highly targeted roles, -but- are extremely unlikely otherwise.
*There's of course more on the data, ML, coding, systems, and domain knowledge sides, but here sticking to the math/stats stuff you asked about
Understand regression, learn ROC curves and what they mean, any good comprehensive course on udemy will cover the math stuff