Probit Regression

Probit regression is used to model binary response variables. It can be used in cases where Logit regression is applicable.

How To Model Probit Regression?

The synax for probit is similar to logit, except for a variation in family parameter.

probitModel <- glm (binaryResponse ~ pred1, pred2, data=inputData, family=binomial(link="probit")) # probit
summary (probitModel) # Model summary
confint (probitModel) # confidence intervals.

As in logistic regression, the co-efficient values should both be positive or negative. If the condidence range includes 0, then it is possible that the coeffient value could take on the value 0, which will effective the same as not including the variable in the model.

predict(myprobit, newData, type = "response", se.fit = TRUE) # predict on test data
logLik(probitModel) # log likelyhood statistic

If you like us, please tell your friends.Share on LinkedInShare on Google+Share on RedditTweet about this on TwitterShare on Facebook