n. in model-building procedures such as backward elimination or stepwise regression, the specific ratio of variances needed to justify keeping a variable as a predictor. Generally, an F-to-remove value around 4.0 is sufficient to allow a variable to be retained in an analysis.