topic OUTPUT of Proc mixed model with Tukey-Kramer adjustment in Statistical Procedures

OUTPUT of Proc mixed model with Tukey-Kramer adjustment

GregTM — Mon, 03 Jul 2023 13:19:55 GMT

Hello,

I'm a novice when it comes to SAS (apologies in advance).

My question really pertains to the difference between values "Pr > |t|" and "adjusted P" columns in the final output of the differences of least squares. It is my limited understanding that both are p-values. But as I now want to highlight significant differences between my treatments, which is the one I should be paying attention to (and why ultimately are there two distinct "p-values")?

Many thanks.

Re: OUTPUT of Proc mixed model with Tukey-Kramer adjustment

sbxkoenk — Mon, 03 Jul 2023 13:44:16 GMT

Hello,

Have you ever heard about inflation of the Type I - error probability due to multiple comparisons (or multiple testing in general)?

That's the problem adjusted p-values try to solve.

A Type I - error means rejecting the null hypothesis when it's actually true. The risk of committing this error is the significance level (alpha or α) you choose. When you do 10 tests, the risk of committing this error at least once is higher than alpha.

When you want to the overall probability (across all comparisons) for a Type I - error to be alpha, you have to bring down the alpha for each test or inflate the p-value for each test.

Hence, adjusted p-values are the ones you need to look at in the 1st place. You need to be "conservative"!

BR,

Koen

Re: OUTPUT of Proc mixed model with Tukey-Kramer adjustment

GregTM — Tue, 04 Jul 2023 08:50:45 GMT

Thank you !