The difference of CI from proc glm and proc means

[HappySASUE]

Could anyone explain what's the difference when calculate the mean and CI of "class" variable in proc means and proc glm. For example I want to calculate CI for 2 levels Y and N, (Note: not the difference between Y and N) I have 2 codes give me 2 different CI
Code 1 :
Proc means data=KK clm alpha=0.05 ;
Var UOSMOLVN;
class lotresp;
Output out=textn n=n mean=mean lclm=lclm uclm=uclm std=std ;
run;

                                                      Output 1:

Analysis Variable : UOSMOLVN Osmolality(mOsm/kg) Numerical ValueLot AEOccurrence(Yes/No) N Obs Lower 95%CL for Mean Upper 95%CL for MeanN 116Y 18

344.1497139	350.8675275
328.3620766	344.0823679

Code 2 :
proc glm data=KK;
class lotresp;
model UOSMOLVN=lotresp;
means lotresp/clm t ;
quit;

                                                       output 2 : 

lotresp N Mean 95% Confidence LimitsNY

116	347.509	344.209	350.808
18	336.222	327.846	344.599

Both codes generate the same means, but why CI are different?

Thanks in advance.

H

[Reeza]

Don’t have time to verify this, but check if GLM is using a pooled SE whereas Means would not be using a pooled SE. 

[PaigeMiller · Accepted solution]

PROC GLM uses the root mean square error from the fitted model to determine confidence intervals, which in other words is the pooled standard deviation across all the values of LOTRESP. PROC MEANS uses the actual standard deviation for each value of LOTRESP. PROC GLM and PROC MEANS do not do the same thing, they're not supposed to do the same thing.