Ordinal Logistic Regression

Ordinal logistic regression, usually the proportional odds model, relates an ordered categorical outcome to predictors by modeling the cumulative log-odds of being at or below each outcome category.

순서형 로지스틱 회귀는 주로 비례오즈모형으로, 각 범주 이하에 속할 누적 로그오즈를 모형화하여 순서가 있는 범주형 결과변수와 예측변수의 관계를 분석합니다.

How do I analyze an ordered categorical outcome such as a Likert rating or severity level?

리커트 척도나 심각도 단계처럼 순서가 있는 범주형 결과변수는 어떻게 분석하나요?

Ordinal logistic regression models an outcome with ordered categories, such as low, medium, and high, by estimating how predictors shift the cumulative odds of higher versus lower categories. The common proportional odds version assumes each predictor has the same effect across all category thresholds, which should be checked with a test such as the Brant test.

When to use Ordinal Logistic Regression

• Single Likert-type rating as the outcome
• Ordered severity, satisfaction, or education levels predicted by several variables
• Ordered outcomes with few categories where treating them as continuous is doubtful

When to use something else

• Unordered categories: use multinomial logistic regression
• Proportional odds assumption is violated: use a partial proportional odds (generalized ordered logit) model

Assumptions and requirements

• Outcome categories have a natural order
• Proportional odds (parallel lines) across thresholds
• Independent observations
• No severe multicollinearity

What the analysis reports

• Coefficients and odds ratios with confidence intervals
• Threshold (cut-point) estimates
• Likelihood ratio chi-square, pseudo R², and AIC
• Test of the proportional odds assumption
• Predicted category probabilities

How to report it

Each one-unit increase in [predictor] was associated with higher odds of being in a higher [outcome] category, OR = [value], 95% CI [[value], [value]], p = [value].

Related methods

• Multinomial Logistic Regression
• Binary Logistic Regression
• Ordinal Probit Regression

Run Ordinal Logistic Regression on Quorant