 |
 |
Hi, I have a problem estimating the regression parameter in a repeated measures analysis with my treatment variable as a linear variable (i.e non class in SAS). I measure my response variable repeatedly at 4 timepoints after baseline. The design of the study is a parallel group study with Placebo as a control and 4 other doses (10,30, 60 and 100 mg). So the model I want to use is a model with baseline as a covariate, dose as a linear effect, time as repeated effect and including the interaction between time and dose and also the time and BL interaction. I am interesting in estimating the regression parameter of dose at each timepoint (i.e. estimating dose*time for each timepoint). I use the following code in SAS: proc mixed data=stat2; class subjid TIME; model y=DOSE BL TIME TIME*DOSE TIME*BL / solution cl e; ods output SolutionF=rlpar; repeated TIME /subject=SUBJID; run; The problem is that in the output gained from using the model option solution I get estimates which gives me estimates that are 0 for timepoint 2.5, so SAS uses timepoint 2.5 as the reference. How do I get the estimates and p-values (for the test that the beta is 0) for each timepoint rather than estimates of the difference between each timepoint and timepoint 2.5? See the example below (fictive results) on how the estimates look like: Effect Time Estimate Intercept . 0.45 BL . -0.32 DOSE . -0.01 TIME 0.75 0.47 TIME 1.25 0.33 TIME 1.75 0.39 TIME 2.5 0.52 DOSE*TIME 0.75 -0.01 DOSE*TIME 1.25 0.01 DOSE*TIME 1.75 0.007 DOSE*TIME 2.5 0 KR, Joakim Englund _________________________________________________________________ Nya Windows 7 gör allt lite enklare. Hitta en dator som passar dig! http://windows.microsoft.com/shop