06-09-2017 11:23 AM - last edited on 06-09-2017 11:27 AM by Reeza
I have a repeated measurement RCT with wait-list control group. All subjects had received identical intervention except wait-listed subjects had 4-week delay before they receive intervention at T2w (See Figure 1). Given my data has missing values and hierarchical (within subject and between subject variations) as well as timing to tests is not consistent across intervention and wait-listed controls, mixed model with autoregressive correlation function would offer a good fit (See Mixed Model code below). Few words about the trial protocol: intervention means 4 different component of treatments intended for intoxication, outcome variables are Physical and Mental Component Scale Score (PCS and MCS) (see attached data "first10"). I will test lab test outcomes later at some point but now I'm analyzing questionnaire based life-quality measures (PCS and MCS) for outcome.
Would you please kindly speak out of your experience on following aspects? I truly appreciate your insights <3
1. Would you agree the way I'm coding "Time" variable shown in Figure 2? It basically is referring to the number of weeks until subjects get an intervention. This idea was inspired by David Howell (See the link for the reference talking about coding the time in RCT but not in the context of wait-list control data).
2. Would you please look at ways I'm coding group status? I'm confused here. Am I supposed to code group status (0 vs 1) reflecting on the particpation in the actual intervention (Groupb column in Figure 2) or how we nominated subjects to intervention vs wait-list control (Groupa column in Figure 2)?
Mixed model code (Using SAS 9.4):
Proc Mixed data = have; class group subj time; model PCS = group time group*time; repeated time /subject = subj type = AR(1) rcorr; run;