Calculating predicted probability of event happening and storing the value for each observation - how to?

[Magnus Carstens]

Hi guys,

I'm writing my master's thesis in Political Science and studying the causes leading to autocratic regime breakdown. In that regard i'm trying to construct a measure for the leader's time horizon.

The common measure in the literature is to calculate the predicted probrability of a regime failure in a given year by running a regression on the different variables that are theoretically believed to influence the probability of regime breakdown, and then storing that coefficient in a new variable for each country-year.

Unfortunately, I'm unsure on how I do this in Stata. How do I store different coefficients relating to the probability of a regime breakdown happening in that specific country in that specific year, for example the risk of a regime failure occurring in Mozambique in 1978?

Does anybody know how this procedure is done?

Best regards,
Magnus

edit: grammar

[Joseph Coveney]

help predict

[Magnus Carstens]

Thank you. I'll look into it tomorrow.

[Clyde Schechter]

I would qualify Joseph Coveney's advice slightly. The -predict- command is available after (I think) every Stata estimation command, but what it can predict differs from one command to the next. The advice you will find at -help predict- is fairly generic. As you have a dichotomous outcome, you will likely be using -logit- or -probit- or perhaps a multi-level version of one of those for your modeling. To get information about -predict- specific to the estimation command you will be using, you should use

help logit postestimation

(or substitute -probit- or whatever the actual estimation command is for -logit-) and then look for the blue -predict- link and click there. That will give you detailed information about what -predict- can do for you specifically after that command.

[Magnus Carstens]

Thank you for the response, Clyde. I'm using logit, so I'll look for that in the help feed. Have a nice day.