Dear Statalist,
I have a retrospective data on a cohort and I am trying to assess differences in the event rates between three groups (three different anti-hypertensive treatments). Because it is retrospective data and the treatment was likely to be assigned based on patients'characteristics I fitted an inverse probability weighting regression adjustment. However I am struggling to interpret the output.
ATET (treatment 2 vs 1) -1.8 (p<0.05)
ATET (treatment 3 vs 1) 1.1 ( p=0.873)
The only significant result is the first one. If I understood it correctly I could comment the result as follow: If everybody in the cohort was taking the treatment 2 the average time to the event is estimated to be 1.8 years lower, compared to if the whole cohort was taking the treatment 1. Is that correct? If I was fitting a normal Cox model that would convert into an HR greater than 1 right?
Does the fact that I have 3 treatments change the way I should interpret the ATET?
Many thanks
I have a retrospective data on a cohort and I am trying to assess differences in the event rates between three groups (three different anti-hypertensive treatments). Because it is retrospective data and the treatment was likely to be assigned based on patients'characteristics I fitted an inverse probability weighting regression adjustment. However I am struggling to interpret the output.
ATET (treatment 2 vs 1) -1.8 (p<0.05)
ATET (treatment 3 vs 1) 1.1 ( p=0.873)
The only significant result is the first one. If I understood it correctly I could comment the result as follow: If everybody in the cohort was taking the treatment 2 the average time to the event is estimated to be 1.8 years lower, compared to if the whole cohort was taking the treatment 1. Is that correct? If I was fitting a normal Cox model that would convert into an HR greater than 1 right?
Does the fact that I have 3 treatments change the way I should interpret the ATET?
Many thanks
