**STAT 330: 98-1
**

Assignment 8 (SAS Asst 2) Comments on Answers

- Sometimes people looked up
*F*-critical points in order to carry out tests at specific levels as requested in the text. This is not wrong but also not necessary. If*P*< 0.01 then you can reject the null at the level 0.01. That's why you only need the*P*-value. - A
*P*-value does not measure . After all, you must assume is true to compute*P*. Instead it measures how often, when the null is true, you would expect to see a test statistic as extreme as the one you actually saw. - You need to interpret the output not just hand it in. Your conclusions must be about real world things not just ``I reject .''
- In one question the Tukey intervals all include 0 but the
*F*test of the hypothesis rejects the null. The interpretation is that while we are confident that not all the means are equal we aren't quite sure which ones differ from which; we are unable to name with adequate confidence a specific pair which differ. - If then we
estimate using . The variance of this is
which is simply

Now plug in MSE for and take a square root to get the estimated standard error of and use the interval (estimated standard error).

Mon Mar 23 23:32:29 PST 1998