For every model we evaluate at , we accept that there is a 5% chance that we reject the null hypothesis when the null hypothesis is actually true.
To avoid a large number of false positives
1. Controlling the family-wise error rate
2. Controlling the false discovery rate
Simulation experiments
Case study I: no true positives
Case study II: 50% true positives
ANOVA (Analysis of Variance)
Analysis of Variance is a generalisation of a two-sided two-sample - test to 3 or more samples
先复习了t-test往前翻就行
How serious is the equal variance assumption for two sample t - test?
In particular, quite strange things can happen if the population variances are different
The sample sizes are very different
The general ANOVA decomposition
aov(formula = weight ~ group, data = PlantGrowth)