A situation in logistic regression when the outcome variable can be perfectly predicted by one predictor or a combination of predictors.

A situation in logistic regression when the outcome variable can be perfectly predicted by one predictor or a combination of predictors.

• A. Complete separation
• B. Compound symmetry
• C. Confirmatory factor analysis
• D. Concurrent validity

Answer: (A)

Complete separation happens in logistic regression when the outcome is perfectly predicted by one predictor or a combination of predictors. In practice, some predictor values (or combinations) separate the data so that all observations with those values have the same outcome (all 0s or all 1s). This makes it possible to perfectly discriminate the classes, which drives the maximum likelihood estimates of the coefficients toward infinity and prevents the model from converging; standard errors become unreliable and p-values can be meaningless. The other terms aren’t about this issue: compound symmetry is a covariance structure used in repeated-measures models, confirmatory factor analysis tests a hypothesized factor structure, and concurrent validity concerns how well a measure relates to a criterion measured at the same time. So the situation described is complete separation.

Complete separation happens in logistic regression when the outcome is perfectly predicted by one predictor or a combination of predictors. In practice, some predictor values (or combinations) separate the data so that all observations with those values have the same outcome (all 0s or all 1s). This makes it possible to perfectly discriminate the classes, which drives the maximum likelihood estimates of the coefficients toward infinity and prevents the model from converging; standard errors become unreliable and p-values can be meaningless. The other terms aren’t about this issue: compound symmetry is a covariance structure used in repeated-measures models, confirmatory factor analysis tests a hypothesized factor structure, and concurrent validity concerns how well a measure relates to a criterion measured at the same time. So the situation described is complete separation.