I'd included the logically problematic post treatment factors. It's
only logically problematic if you're putting it into the causal model.
You're not, you're just trying to control for it, so I don't think it
matters.
Rather than statalist, you might try the psych-methods list - that's a
little more friendly, or crossvalidated, which is at
stats.stackexchange.com.
Jeremy
On 31 July 2013 02:52, Iain Brennan <[log in to unmask]> wrote:
> Hi All,
>
> I'm no longer a post-grad, but I'm scared of the orcs on the stata list and I do the online surveys all the time, so I hope you don't mind my hijacking this list briefly.
>
> Using a large, cross-sectional sample of victims of violence, I am interested in testing the effect of alcohol intoxication (let's call it the 'treatment') on victims' ratings of the seriousness of the assault (outcome).
>
> I have run regression analyses to explore factors that affect seriousness ratings and, as you might expect, the amount of injury done is the strongest predictor of seriousness.
>
> Victims who were drunk at the time of their assault had somewhat different characteristics from the other victims (e.g. more men, less educated...) and that that assaults on drunk victims resulted in more serious injury than assaults on sober victims.
>
> I thought it might be a good idea to match victims with just the 'drunk at time of assault' as a distinguishing factor ('treatment') so that I can compare (drunk) apples with (sober) apples.
>
> I have run propensity score matching to identify a matched sample of sober and drunk victims. My problem is that, because important confounding factors like injury happened after the 'treatment allocation', I couldn't logically include those in my matching model. So, when I compare the groups, the confounding effect of injury still remains.
>
> As far as I can see, I'm faced with three options but I'm probably missing something:
>
> 1. Include the statistically relevant, but logically problematic, 'post-treatment' factors in the matching.
> 2. Control for the propensity score and the post-treatment factors in a regression of the outcome on victim intoxication using just the matched sample.
> 3. Stop messing around with treatment evaluation on a cross-sectional data set and just stick with regression.
>
> Any help on this would be greatly appreciated.
>
> Iain
>
> Lecturer in Criminology & Psychology,
> University of Hull
