Logistic Regression vs. Probit Regression

[prakritinaswa]

What are the key differences between Logistic and Probit regression?
How does one decide which of them is relevant to a problem?

 

[maartenbuis]

A probit will approach 0 and 1 slightly faster than a logit, but the real answer is that the two are virtually the same. In practice you would not be able to choose between them based on real data. So, you just pick whichever one you find most convenient. I prefer the logit, as the pareameters have actual meaning.

 

[Dason]

The parameters in a probit can have meaning too.

The main 'advantage' I see to the logit is that the odds ratio is constant which makes for interpreting things a little bit easier.

 

[Dason]

We actually had a decent sized thread that explored some of these ideas here: http://www.talkstats.com/showthread.php/20004-Logit-v.-Probit-A-fight-to-the-death