Stratified Probability Testing

For our university sample, we would probably want to make sure that the 500 person sample had relatively the same percentage of freshmen, sophomores, juniors, and seniors as did the university. To assure this, we would actually take four samples-one of each class from the registrar's list-and then take sufficient names from each class. Suppose that the composition of the university was such that:

30% of students were freshmen;
27% were sophomores;
23% were juniors;
20% were seniors.

We would therefore want a sample with:

150 freshman (30% of 500)
135 sophomores (27% of 500)
115 juniors (23% of 500)
100 seniors (20% of 500).

We could do this by selecting 150 names from the freshman hat, 135 from the sophomore hat, 115 from the junior hat, and 100 from the senior hat.

Again, the hat is not necessary; computers can easily be programmed to do this.