STAT 301 Lecture Notes - Lecture 19: Family-Wise Error Rate, Simple Linear Regression, Type I And Type Ii Errors
Document Summary
If we run 100 experiments in which h0 is true, then using an alpha level of 0. 05 we expect to mistakenly reject h0 5 times. The probability of a type i error for each individual comparison. This is typically an alpha level of 0. 05. The probability that at least one type i error is committed in a family of comparisons. If we want a 5% family-wise error rate, we will have to use a smaller comparison-wise error rate. This will mean that p-values need to be smaller to be called significant, and confidence intervals will become wider. Controlling the family wise error rate by reducing the comparison wise error rate is a method of correcting for multiple comparisons. There are many different kinds of multiple comparisons corrections. The most popular multiple comparisons correction is the bonferroni correction. If two variables are correlated, then they move together. As one increases, the other one either tends to increase or tends to decrease.
