Multinomial Logistic Regression

Multinomial logistic regression is a regression model for an unordered categorical outcome with three or more categories that estimates the log-odds of each category relative to a reference category.

다항 로지스틱 회귀는 순서가 없는 세 개 이상의 범주를 갖는 결과변수에 대해 각 범주의 기준범주 대비 로그오즈를 추정하는 회귀모형이다.

How do I predict an outcome with three or more unordered categories?

순서가 없는 세 개 이상의 범주로 된 결과를 어떻게 예측하나요?

Multinomial logistic regression models an unordered categorical outcome with three or more categories by estimating log-odds for each category against a reference category. It fits choices such as transport mode or brand, while ordinal logistic regression is used when categories have a natural order.

When to use Multinomial Logistic Regression

• A nominal outcome with three or more categories, such as brand choice or occupation type
• Continuous or categorical predictors of category membership
• Interpreting relative risk ratios against a reference category

When to use something else

• Ordered categories; use ordinal logistic regression
• Exactly two categories; use binary logistic regression

Assumptions and requirements

• Independent observations
• Independence of irrelevant alternatives (IIA)
• No severe multicollinearity
• Adequate cases in each outcome category

What the analysis reports

• Coefficients and relative risk ratios for each category versus the reference
• Likelihood ratio test of the model
• Pseudo R-squared
• Classification table

How to report it

Relative to [reference category], each unit increase in [predictor] was associated with higher odds of choosing [category], RRR = [value], 95% CI [[value], [value]], p = [value].

Related methods

• Binary Logistic Regression
• Ordinal Logistic Regression
• Conditional Logit Model

Run Multinomial Logistic Regression on Quorant