You need to determine whether a variable should be removed in the stepwise variable selection methods in logistic regression. Which two metrics are used to accomplish this task? (Choose two.)

Question

Options

• AWald Statistic
• BKS-Statistic
• CLikelihood Ratio (LR) Statistic
• DZ-Statistic

• A
  75% (39)
• B
  10% (5)
• C
  15% (8)

Explanation

In stepwise logistic regression, variable removal decisions are made at the individual coefficient level - specifically, testing whether a coefficient is statistically indistinguishable from zero. The Wald Statistic (A) does exactly this by computing (β/SE(β))², and the Z-Statistic (D) is its square root (β/SE(β)), making them two expressions of the same underlying test for a single variable's significance. The KS-Statistic (B) is a model-level discrimination metric (measuring separation between event/non-event distributions), not a variable-selection tool. The Likelihood Ratio Statistic (C) compares two competing nested models for overall fit and is better suited for model-level comparisons than for per-variable removal decisions in stepwise procedures.

Memory tip: "Wald and Z test whether to Zero out a variable" - both check if a single coefficient equals zero, which is precisely the criterion for dropping it from the model.

No community discussion yet for this question.