[nashke]

Alcohol is strongly associated with suicidality, but again, it's difficult to prove causality (Try randomizing people to "alcohol binging group" or "suicidal ideation group"). The association is more robust than that for cannabis. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2872355/

[vintermann]

Do you know if there have been any mendelian randomization studies looking at alcohol and suicidality?

[nashke]

None that I know of, but I am not as updated on research of alcohol use disorders as I am with cannabis. One interesting frontier is using machine learning methods for causal inference (the reason I have some vague knowledge of this is that my brother is working on this...). These may hopefully allow for some interesting analyses of existing data. Here is a link for his seminar on causal inference for observational studies: http://www.cs.nyu.edu/~shalit/slides.pdf