Two-way ANOVA with Dunnett to get P-value

[lizzy28]

Hi Everyone,

I'm trying to do two-way ANOVA check with Dunnett option but need to have p-value for the mean comparison. I use the code below:

proc anova data=dataset;

class A B;

model COST = A | B;

 means A | B /dunnett; 

run;

The last piece of the test has the following table:

I need to have P-values to see if means are significantly different across the groups. Anyone knows how to do it?

Thanks a lot!

Lizi

[PaigeMiller]

I don't think you can perform a Dunnett's test on an interaction in SAS. I believe it only applies to main effects.

I think you should not use PROC ANOVA here, you really ought to use PROC GLM to analyze this data (although you still can't do Dunnett's test on interactions in either PROC), but GLM will probably provide more accurate results. From the PROC ANOVA documentation:

Caution: If you use PROC ANOVA for analysis of unbalanced data, you must assume responsibility for the validity of the results.

[PaigeMiller · Accepted solution]

Adding

If you want to do Dunnett's test on interactions, you can do this with Least-Squares Means (which you should probably be using anyway with this unbalanced data). 

http://documentation.sas.com/?cdcId=pgmmvacdc&cdcVersion=9.4&docsetId=statug&docsetTarget=statug_glm...