Is it possible to analyze the effects of an independent binary variable in regression via average treatment effect (ATE) while the binary variable of interest is a certain characteristics of individuals in a survey (for example risk aversion or gender)?
It seems that ATE can be only used when there are two distinct groups that are treated and not treated for a certain thing. But in my case, there is no treatment, rather the differences in demographic characteristics of individuals.
Hello Omid,
Average treatment effect (ATE) is usually reserved for quantifying the mean difference on some metric outcome variable between two randomly-formed batches (no-treatment, treatment). You may certainly estimate and report the predicted/estimated unit DV difference for between dichotomous levels for a given IV (holding all others constant).
Whether folks would be happy with your calling that an ATE (when the groups are non-randomized, and therefore any apparent difference could be confounded with one or more other variables not included in your model) is another matter. So, why not just call it what it is? Removing the ATE label doesn't change your ability to interpret relative impact of variables (at least, relative to those that were included in the model) for your sample.
Good luck with your work.
Hi Omid!
I agree with David. You do have great expertise on these topic within your own University. I suggest you talk to:
Michael Lechner
- Badges
- Science topic
- Similar topics
- Statistics
- Survey