Hi, I am analyzing systolic blood pressure (SBP) variation over time among subjects under the intervention. The response variable is SBP measured weekly starting from baseline over the 24 week period (1 - 24weeks). I have missing data here and there in those measurments. Subject were categorized into 3 groups based on their baseline SBP values ( (SBPgr_bsl = 1 if normotensives; 2 - with moderate hypertension and 3 - with severe hypertension ) . The covariates are sex, age (in years), diagnosis (copd, heart failure). The exploratory analysis of SBP variation over time showed steeper reduction at earlier weeks during the study so I create an additional logarithmically transformed "week" variable (log_week) to fit the data. I am looking for the appropriate command for the analysis, here are the options
Proc mixed data= monitoring;
class SBPgr_bsl diagnosis sex week ; /* I keep "week" variable here to be able to define it under repeated statement*/
model sbp = SBPgr_bsl diagnosis sex age log_week SBPgr_bsl*log_week/solution;
repeated week /type=ar(1) sub= id_patient r rcorr;
run;
or
Proc mixed data= monitoring;
class SBPgr_bsl diagnosis sex ;
model sbp = SBPgr_bsl diagnosis sex age log_week SBPgr_bsl*log_week/solution;
random int log_week /type=un sub= id_patient;
run;
or
Proc mixed data= monitoring;
class SBPgr_bsl diagnosis sex week ;
model sbp = SBPgr_bsl diagnosis sex age log_week SBPgr_bsl*log_week/solution;
random int /sub= id_patient;
repeated week /type=SP(pow)(log_week) sub= id_patient rcorr;
run;
Also i would like to report on differences between baseline and 24th week among SBPgr_bsl = 1, SBPgr_bsl = 2 SBPgr_bsl = 3, however lsmeans statement is working when i have time as a categorical variable and not as a continous. Would you please advise how to request the above mentioned differences?
Thanks you, Yeva