Why this matters
Anyone can run a model in a few lines of code. What employers pay a data scientist for is judgement: knowing how sure you are, knowing when a difference is real, and knowing when a pattern is only a coincidence. That judgement comes from statistics. It is also the skill that protects people. When an NGO decides which communities get a health programme, or a lender decides who gets credit, a careless statistical claim has real consequences.
The big idea: you almost never see everything
You rarely have data on every farmer, customer or household. You have a sample, and the sample would have come out a little differently if you had picked other people. Statistics is the toolkit for reasoning about that variation honestly. Every lesson in this section answers a version of one question: given the data I have, what can I fairly say about the wider world, and how sure can I be?
What this section covers
- Probability and distributions: how random variation behaves, what common shapes of data look like, and why averages from samples settle into a predictable pattern.
- Sampling, confidence and hypothesis tests: confidence intervals, p values, and the limits of both, so you neither overclaim nor miss real effects.
- Experiments and A/B tests: why random assignment is the most reliable way to show that a change caused a result, how big a test needs to be, and the mistakes that spoil it.