This is my short introduction to statistics. It is published under Creative Commons license CC BY-NC. You can use this work non-commercially, but credit must be given, and there is no allowance for commercial use.
I designed and built this course for One Acre Fund to quickly train their data analysts and data scientists in basic statistical concepts that are vital to trial design and analysis. Due to the atypical background of One Acre Fund analysts, this course is designed to be as math-lite as possible. The aim of this course is therefore to build a kind of statistical intuition and statistical critical thinking.
At the end of this course, the reader should be able to oversee, design, and analyse a variety of trials important in impact-evaluation and product innovation.
Highlights:
- Interpreting and plotting distributions
- Null hypothesis testing, p-values and significance
- Non-parametric hypothesis testing
- Power and sample size calculations
- Monte-Carlo methods for non-parametric power calculations
Highlights:
- Formulating effective hypotheses
- Randomization
- Intra-cluster correlation and sample size
- P-value thresholds (aka alpha levels)
- Using minimum detectable effects in sample size calculations
Highlights:
- Analysing cluster randomized trials
- Summarizing clusters
- Cluster-robust methods with mixed-effect models
- Diagnosing and interpreting regression methods
Highlights:
- Analysing binary outcome variables
- Fixed effect models
- Mixed effect logistic modelling
Highlights:
- Introducing the multiple comparison problem (and corrections)
- Introducing positive predictive value and other advanced thoughts on power
- Refreshing some useful R functions