Hi all. I apologize in advance if the questions will sound a bit too theoretical and not software related, but I could use a little help from you all.
I fitted a logistic regression model with a binary exposure variable X, trying to understand the effect it has on a given outcome Y, which measures infection (yes/no). I calculated the odds ratio of the variable of interest as well as the Average Marginal Effect, via the margins command, both highly significant. Particularly, in my case the AME associated with X=1 is -0.0245 [-0.0201;-0.0283]. This means that the probability that Y=1 (infection occurs) is reduced by 2.45% when X=1.
Now, how do I move from this to a simpler explanation regarding the raw count difference between the two groups (i.e., treated and control)? Something like:
"given that in my dataset I have 2,000 observations, with 1100 treated and 900 untreated, the AME is telling us that the probability that Y occurs is 2.45% lower for treated, translating to a reduction of Z infected patients"
What do I have to use? The baseline unconditional probability of infection overall? The baseline for both groups? I am a little bit confused about the process. Thanks for any help!
I fitted a logistic regression model with a binary exposure variable X, trying to understand the effect it has on a given outcome Y, which measures infection (yes/no). I calculated the odds ratio of the variable of interest as well as the Average Marginal Effect, via the margins command, both highly significant. Particularly, in my case the AME associated with X=1 is -0.0245 [-0.0201;-0.0283]. This means that the probability that Y=1 (infection occurs) is reduced by 2.45% when X=1.
Now, how do I move from this to a simpler explanation regarding the raw count difference between the two groups (i.e., treated and control)? Something like:
"given that in my dataset I have 2,000 observations, with 1100 treated and 900 untreated, the AME is telling us that the probability that Y occurs is 2.45% lower for treated, translating to a reduction of Z infected patients"
What do I have to use? The baseline unconditional probability of infection overall? The baseline for both groups? I am a little bit confused about the process. Thanks for any help!
