Probit Regression Model

Probit regression is a model for a binary outcome that links the probability of the event to predictors through the cumulative standard normal distribution.

프로빗 회귀는 사건이 일어날 확률을 표준정규 누적분포함수를 통해 예측변수와 연결하는 이분형 결과변수 모형이다.

How do I model a yes/no outcome assuming an underlying normally distributed latent variable?

정규분포를 따르는 잠재변수를 가정하고 예/아니오 결과를 어떻게 모형화하나요?

Probit regression models a binary outcome through the standard normal cumulative distribution, which corresponds to a normally distributed latent propensity. It gives conclusions similar to logistic regression and is common in economics, usually interpreted through marginal effects.

When to use Probit Regression Model

• A binary outcome such as participation, purchase or default
• A latent-normal interpretation is theoretically motivated
• Part of models that assume normal errors, such as Heckman selection or bivariate probit

When to use something else

• Odds ratios are wanted for interpretation; use logistic regression
• Ordered outcome categories; use ordered probit or ordinal logistic regression

Assumptions and requirements

• Independent observations
• Binary outcome
• Correct link and linear predictor specification
• No perfect separation

What the analysis reports

• Probit coefficients with standard errors
• Average marginal effects
• Likelihood ratio chi-square
• Pseudo R-squared

How to report it

A probit model showed that [predictor] increased the probability of [outcome], average marginal effect = [value], z = [value], p = [value].

Related methods

• Binary Logistic Regression
• Ordered Probit Model
• Heckman Selection Model

Run Probit Regression Model on Quorant