Probit or LPM?

[Oliver Edward Webster]

Hi Statalist,

I am researching the wage and employment differences between graduates who have and have not attended a Russell Group University. As part of this I have undertaken an audit study sending out comparable CVs and observing the call back rate. I am using my results to look at how Russell Group status, a binary variable, affects the probability of a callback (another binary variable) as a result I am debating whether or not to use a probit model or linear probability model.

My main query is how would I be able to get this output from Stata?. Can I even get output from such statistical models given the binary nature of the independent variable?

Many Thanks

[Richard Williams]

Unless I misunderstant the problem, you can just use the regress or probit commands (or, if you prefer, logit). I am not a fan of the lpm but if the model is as simple as you make it sound then it won't matter much. It sounds like you just want to analyze a 2 by 2 table, in which case the prtest or tab2 commands would do.

If the problem is more complicated than it seems to me, then please elaborate. I'm not clear on what it is that makes you unsure of what to do. Binary dependent and independent variables are certainly not hard for Stata to handle.

[Oliver Edward Webster]

Yes you put it in much clearer terms, I am looking to regress a 2x2 table. So just to follow from what you advised, should I put my data in to the data editor then use the command tab2[variablenames] to create such a table. How do I launch a probit or LPM from that table?
(sorry if this is painfully basic)


further from that regression I may add other variables such as Sector or proposed salary, how could one launch a regression to examine the effect of attending Russell Group unis on call back rates when it is beyond a 2x2 table?