²ÊºçƵµÀ

Binomial Probability Test

When you have variables with values like yes/no or fail/succeed, you're working with a binomial variable. If you expect a certain distribution of that binomial variable, you can test it using the bitest command. This may sound confusing in the abstract, so lets work with an example:

Say I am curious if ²ÊºçƵµÀ's population is better at spatial reasoning than a general US sample - a completely made-up meta-analysis reveals that 38% of the US population passes spatial reasoning tasks. I subject a random sample of 50 ²ÊºçƵµÀ students to the same test. The variable is called "spatial" and the command to see if ²ÊºçƵµÀ passes the test more often than the general US population would be bitest spatial=.38

The result of my test is highly significant whether I use the better performance hypothesis Pr (> = 40) or the two-tail test Pr(k <= 1 or k >= 40). This means that more ²ÊºçƵµÀies succeed at a spatial reasoning test than the general poluation.