Proc GLM: Which model?

ChrisMac — Mon, 20 Jan 2020 22:33:44 GMT

Hello

I was part of a randomized study in which patients received active drug or placebo (TREATMENT). The hematocrit was measured at baseline (BASELINE_HCT) and after 8 mo. (FOLLOW_UP_HCT).

I received the following code from a colleague, but I have difficulties in finding out what kind of model this is. Is it a generalized linear model where the least-squares mean is calculated?

Thanks in advance.

Chris

proc glm data=DATASET;

class TREATMENT;

model FOLLOW_UP_HCT=TREATMENT BASELINE_HCT;

lsmeans TREATMENT /pdiff stderr cl;

run;

Re: Proc GLM: Which model?

PaigeMiller — Mon, 20 Jan 2020 23:17:30 GMT

I have to say I am troubled by the absence of patient in this model. It seems to me that a patient effect is needed here.

I also think that if you are trying to see which treatment results in the largest change in hematocrit level, you might want to treat this as a "paired" experiment rather than having FOLLOW_UP_HCT and BASELINE_HCT as separate terms in the model, you might want to look at the delta between followup and baseline as the response variable. Or you might want to see if the effectiveness of treatment depends on BASELINE_HCT, that would be different and require a different model than the one you show.

Those are the questions you ought to be asking. What the model is named is somewhat irrelevant, you will have to justify you have the right model.

topic Re: Proc GLM: Which model? in Statistical Procedures

Proc GLM: Which model?

Re: Proc GLM: Which model?