Day 05 Part 2 — Statistics for Analytics: Agenda¶
Statistics is the language of uncertainty. Analysts without statistics give confident answers to uncertain questions. With statistics, you can quantify how confident you should be, detect real changes vs. random noise, and design experiments that produce trustworthy results.
Session Overview¶
Duration: 3 hours Prerequisite: EDA — concepts like mean, median, and distribution should be familiar Tools: Python (scipy, numpy) in Jupyter Lab
Learning Objectives¶
By the end of this session you will be able to:
- Explain and compute central tendency (mean, median, mode) and understand when each is appropriate
- Explain variance and standard deviation in business terms
- Understand the normal distribution and the Central Limit Theorem
- Calculate and interpret confidence intervals
- Run a t-test and interpret the p-value
- Design and analyse a simple A/B test
- Understand common statistical mistakes analysts make
Session Flow¶
| Time | Topic | File |
|---|---|---|
| 0:00 – 0:25 | Mean, median, mode — the right measure for the right data | 01-mean-median-mode |
| 0:25 – 0:50 | Variance, standard deviation — measuring spread | 02-variance-and-std |
| 0:50 – 1:20 | Probability basics — the foundation of inference | 03-probability-basics |
| 1:20 – 1:50 | Distributions — normal, binomial, and their business uses | 04-distributions |
| 1:50 – 2:30 | Statistics for business — hypothesis testing and A/B tests | 05-statistics-for-business |
| 2:30 – 3:00 | Practice and review | 06-practice-questions |
Why Statistics Matters for Analysts¶
| Situation | Without statistics | With statistics |
|---|---|---|
| Revenue up 5% MoM | "Revenue grew!" | "Revenue grew 5%, but with high variance this could be random — p=0.12, not significant" |
| A/B test: Button A vs B | "B had more clicks" | "B had 8% more clicks (95% CI: 4-12%), p=0.003, statistically significant" |
| Average order £189 | Report the average | Report median (£147) — mean is inflated by 3 whale customers |
| New feature launched | "Conversion up 3% after launch" | Control for seasonality, run a proper test, avoid confounders |
← Visualisation Interview Questions · Next: Mean, Median, Mode →