topic Re: Reporting T-values in LSMEANS Statement in PROC GLM in Statistical Procedures

Reporting T-values in LSMEANS Statement in PROC GLM

Nebulus — Sun, 21 Jun 2015 02:37:44 GMT

I'm trying to get SAS to spit out t-values for my contrasts in LSMEANS. Is there any way to achieve this? It just reports p-values for each comparison.

My code is

LSMEANS species*wax / pdiff stderr;

Re: Reporting T-values in LSMEANS Statement in PROC GLM

Rick_SAS — Mon, 22 Jun 2015 14:22:08 GMT

In general, if p is a probability for a t-distributed statistic, then you can recover the t value by using the QUANTILE function. You can use the OUT= option on the LSMEANS statement to produce an output data set, or you can use ODS OUTPUT.

However, you have to know the degrees of freedom for the pairwise comparison and the significance level of the test. A discussion with formulas is in the doc: SAS/STAT(R) 13.1 User's Guide

I am not an expert in multiple comparisons, but I know that they can be tricky, which might be why the procedure does not output the degrees of freedom.

Re: Reporting T-values in LSMEANS Statement in PROC GLM

lvm — Mon, 22 Jun 2015 15:34:23 GMT

I think you should switch to PROC MIXED (or GLIMMIX). Much better functionality, and you get the t values and df (using many different possible df methods).

Re: Reporting T-values in LSMEANS Statement in PROC GLM

SteveDenham — Mon, 22 Jun 2015 17:36:05 GMT

I agree with , but if you must use GLM, make the following your LSMEANS statement:

LSMEANS species*wax /tdiff pdiff stderr;

Steve Denham

Re: Reporting T-values in LSMEANS Statement in PROC GLM

Nebulus — Thu, 25 Jun 2015 23:22:57 GMT

Thanks all.