Problem in Display of P-value : Proc Glm adjust = dunnett

[deleted_user]

Hi all,

I am facing a problem with Dunnett`s multiple comparisons test with a reference/control group.

Here is the code :

ods output lsmeans = _tpvalues;
proc glm data=_talldata_wide;
title 'Pair-wise comparison with Dunnett correction';
class _pooled ;
model _L21 _L22 = _pooled ;
lsmeans _pooled /pdiff = control ("0") adjust=Dunnett;
run;
quit;
ods trace off;



_pooled is the variable that contains the dose group codes, but control group coded as 0 than its actual value.

For Eg :

dosegrpc _pooled
------------- ------------
5 0 /*coded as 0 since its Control Group*/
15 15 /* treated group */
50 50 /* treated group */


Now say you have one control group and one treated Group

_pooled
------------
0
15


The above proc glm, produces a P-value against dose_group_code = 0, rather than against 0. This is the problem.

Can anybody suggest what is going wrong here. Your help would be greatly appreciated !

Thanks,

Neeral Beladia

[1162]

I'm not sure I understand what you are trying to achieve, but I understand what's happening in your code. You've identified _pooled as a class variable. The values in this variable will be treated as categorical rather than continuous.

If you want these values treated as numeric values instead of categories, you should remove the class statement. If you want to use the original values (5, 15, 50), then you should use the dosegrpc variable in your proc glm.

Hope this helps.