How to get standard deviation from proc glm

[haoduonge]

Hi all,

Output from [proc glm] includes [estimate and standard error], can we get standard deviation instead?

Thanks!

Hao

https://communities.sas.com/t5/ODS-and-Base-Reporting/output-tables-from-proc-glm/m-p/239152 

[PaigeMiller]

Standard deviation of what?

[haoduonge]

Standard deviation of the estimate (continuous variable).

[PaigeMiller]

Ok, estimate of what??

[haoduonge]

Below is the example:

Proc glm data=temp1;

class age_group(ref=first);

model BMI=age_group/solution E3;

run; 

Thanks!

[PaigeMiller]

This GLM model produces a lot of estimates. Which one are you referring to?

[haoduonge]

I think only one estimate (parameter), the value below intercept.

[PaigeMiller]

So to make the example concrete, let's go with 

proc glm data=sashelp.class;
    class sex;
    model height=sex/solution;
    means sex;
run;
quit;

this produces

and so you want the standard deviation of the estimate -3.3211111 for Sex F.

As far as I understand things, there is no standard deviation of this estimate. In fact, I don't even think it makes sense to talk about a standard deviation of this estimate. The Standard Error of 2.2862... gives you a measure of how variable this estimate is. 

So, under the usual assumptions for PROC GLM (errors are iid normal), this estimate has a normal distribtuion with mean of -3.3211111 and a standard deviation — which is equal to the standard error — of 2.2862 ...). Is that what you want?

[haoduonge]

Yes!