In logistic regression, what does rejecting the null hypothesis for a predictor's coefficient indicate?

In logistic regression, what does rejecting the null hypothesis for a predictor's coefficient indicate?

• A. Rejecting the null indicates the predictor significantly contributes to predicting the outcome.
• B. Rejecting the null means the predictor is not significant.
• C. P-values assess overall model fit.
• D. Rejecting the null determines sample size required.

Answer: (C)

In logistic regression, each predictor has a null hypothesis that its coefficient is zero, meaning that predictor has no association with the outcome. Rejecting that null means there is statistically significant evidence that the predictor does influence the outcome; the coefficient is not zero, so the predictor contributes to predicting the probability of the event. The sign of the coefficient shows the direction of that association, and taking the exponent gives the odds ratio for a one-unit change in the predictor. P-values for individual coefficients assess that specific effect, not the overall model fit. If you want to judge the model as a whole, you’d use a global test (like a likelihood ratio test) or model fit metrics (AIC, BIC, pseudo-R^2), rather than interpreting a single coefficient’s p-value as the model’s overall adequacy.

In logistic regression, each predictor has a null hypothesis that its coefficient is zero, meaning that predictor has no association with the outcome. Rejecting that null means there is statistically significant evidence that the predictor does influence the outcome; the coefficient is not zero, so the predictor contributes to predicting the probability of the event. The sign of the coefficient shows the direction of that association, and taking the exponent gives the odds ratio for a one-unit change in the predictor.

P-values for individual coefficients assess that specific effect, not the overall model fit. If you want to judge the model as a whole, you’d use a global test (like a likelihood ratio test) or model fit metrics (AIC, BIC, pseudo-R^2), rather than interpreting a single coefficient’s p-value as the model’s overall adequacy.