have a simple question about using "proc mixed" for repeated measurement.For example, my simple data set is like this --> (this is a made-up dataset, purely for illustration purposes)
SALES VOLUME and Promotion Spending for 8 consecutive months for 3 stores.
data my_sales; input store_id $ x_volume x_promotion mnth; cards; A 200 10 1 A 250 10 2 A 270 12 3 A 200 10 4 A 300 20 5 A 600 30 6 B 290 12 8 B 400 22 1 B 250 10 2 B 270 12 3 B 200 10 4 B 300 20 5 B 600 30 6 B 290 12 8 C 290 12 8 C 405 24 1 C 259 10 2 C 277 12 3 C 201 13 4 C 340 22 5 C 604 33 6 C 297 12 8 ; RUN;
I would like to use "compound symetry" structure for covariance to take account of repeated measurement error. That is, within a subject, error correlates over time. However, I do not believe month sequence really matters. Therefore I do not want to include 'mnth' in the model.
But if I use this ---->
Proc mixed data=my_sales; class store_id; model x_volume = x_promotion /s; repeated / type=cs subject=store_id; run;
Does it mean that repeated measurement over time has been taken care of?
--------------------------------------------------------------------------- Or, shall I use this instead -->
Proc mixed data=my_sales; class store_id; model x_volume = x_promotion mnth /s; repeated / type=cs subject=store_id; run;
????
Your help is greatly appreciated.