How to perform ancova using proc mixed - step by step

[gkmzinsou]

Hi,

I am having a hard time understanding the ANCOVA model and how to implement it in SAS.

I am supposed to do an ANCOVA to obtain adjusted means and adjusted means differences for Y = change from baseline using treatment as a factor variable (3 categories), country as a categorical covariate (5 categories) and baseline as a continuous variable (hence the need for ANCOVA), the last two being adjustment variables. 

I have to use proc mixed and not proc glm or proc reg (Although I would love to know how it's done using these models as well).

My SAS code looks something like:

proc mixed data = mydata;

ods output lsmeans = ls_means;

class trt country;

model change = trt country baseline;

lsmeans trt / pdiff cov ;

run;

Is this correct? I also learned that an ANCOVA model requires the centering of the continuous variable. Is this something that has to be done manually or is it automatically done through SAS? I read somewhere that SPSS and R do it automatically. I know that the ANCOVA model looks something like this:

How do I write my model in this form(knowing that I have an extra categorical adjustment variable)? Or can I just write it as a regression model:

Change = Bo + B1 treat1 + B2 treat2 + B3 country1 + B4 country2 + B5 country3 + B6 country4 + B7 baseline + error (treat 3 = ref and country5 = ref)?

One last thing : I am supposed to do a sequential comparison. 

Treat 1 vs Treat 3

Treat 2 vs Treat 3

Treat 1 vs Treat 2

How do I do that in SAS? Do I use the estimate options? I'm not too sure how that works either. 

Any and all help would be appreciated!

[PaigeMiller]

I think the code you provided works for ANCOVA.

I don't think centering of the continuous variable is required; SAS does not do this automatically.

The PDIFF option of LSMEANS gets you the comparisons of the levels of TRT; there's no need to use LSMESTIMATE for comparing one level to another.