Could a difference this big
be nothing but chance?
Statistics answers one question. These twenty lessons take that sentence apart — what chance is, how you measure this big, which ruler to use, and how the answer fools you.
The examples are not invented. They are this site's own computed results — the mutation burden of 566 TCGA-LUAD patients, 347 miRNA tests, and twelve tumour types across 5,202 slides. Python runs right inside this page.
How you'll learn
Null distributions were re-explained from scratch in /viper and /diggit, multiplicity in /mirna and /til. This course lays that foundation once.
Not dice and factory defects but this site's own computed results — median 6.68 against mean 10.02, 103 out of 347, three out of twelve.
Instead of memorising formulas you shuffle a null distribution into existence and bootstrap a confidence interval yourself. Then the formula stops being a black box.
The last five lessons are the cases where the arithmetic is right and the sentence is wrong — the winner's curse, Simpson's paradox, and the trap inside the word 'adjusted'.