Analyses

Probit regression

Predict a yes-or-no outcome with a probit link instead of odds.

Edit on GitHub

When to use it

Use probit regression for the same binary outcome as logistic regression when you want the slope on the probit scale. A positive slope still means the predictor makes a “yes” more likely. There is no odds ratio on this table.

Assumptions

The outcome is binary. The underlying probit model is a fair summary, and the rows are independent. Perfect separation makes the slopes unstable, the same way it does for logistic regression.

Running it in Tensr

Analyze → Regression → Probit. In chat: “Probit regression of passed on practice.”
Dependent is the 0/1 outcome, unless derive_binary_from builds it. Independents need at least one predictor.
Confidence level starts at 0.95. Include constant is on the request and is not forwarded, so the fit keeps an intercept.

Options

Prop

Type

Reading the output

The same pass/fail outcome and the same hours predictor, fit as a probit instead of a logit.

TermEstimatep-value
const-0.96.085
hours0.187.085
PredictorMarginal effect (dP/dX at mean)
hours0.075

Probit regression predicting passed from 1 predictor(s), n = 96. The primary result is not significant (p = .085). This is not large enough to treat the comparison this page is about as a reliable association. The result is non-significant: the data are still compatible with no effect. Metrics: Pseudo R² = 0.023; Model χ² = 3.044; Model p = .081; AIC = 133.999; Observations = 96.

Reporting (APA 7)

Probit regression predicting passed from 1 predictor(s), n = 96. This result is not significant (p = .085). Report the estimate with that p, and do not describe the pattern as a reliable effect.

Coming from SPSS

Analyze → Regression → Probit.

SPSS Probit is often aimed at dose-response tables. Tensr fits a binary probit from row-level 0/1 data and a numeric predictor.

Odds ratios for the same kind of outcome are on logistic regression.