Augmented Backward Elimination: A Pragmatic and Purposeful Way to Develop Statistical Models

Abstract

Statistical models are simple mathematical rules derived from empirical data describing the association between an outcome and several explanatory variables. In a typical modeling situation statistical analysis often involves a large number of potential explanatory variables and frequently only partial subject-matter knowledge is available. Therefore, selecting the most suitable variables for a model in an objective and practical manner is usually a non-trivial task. We briefly revisit the purposeful variable selection procedure suggested by Hosmer and Lemeshow which combines significance and change-in-estimate criteria for variable selection and critically discuss the change-in-estimate criterion. We show that using a significance-based threshold for the change-in-estimate criterion reduces to a simple significance-based selection of variables, as if the change-in-estimate criterion is not considered at all. Various extensions to the purposeful variable selection procedure are suggested. We propose to use backward elimination augmented with a standardized change-in-estimate criterion on the quantity of interest usually reported and interpreted in a model for variable selection. Augmented backward elimination has been implemented in a SAS macro for linear, logistic and Cox proportional hazards regression. The algorithm and its implementation were evaluated by means of a simulation study. Augmented backward elimination tends to select larger models than backward elimination and approximates the unselected model up to negligible differences in point estimates of the regression coefficients. On average, regression coefficients obtained after applying augmented backward elimination were less biased relative to the coefficients of correctly specified models than after backward elimination. In summary, we propose augmented backward elimination as a reproducible variable selection algorithm that gives the analyst more flexibility in adopting model selection to a specific statistical modeling situation.

High tacrolimus clearance - a risk factor for development of interstitial fibrosis and tubular atrophy in the transplanted kidney: a retrospective single-center cohort study

“…During the model building, covariates which had P ‐values above 0.2 or changed the other odds ratio estimates more than 0.05 were eliminated from the model in a stepwise fashion as described by Dunkler et al . . The same covariates were included in initial model 1 and 2: Continuous estimated clearance was included as a passive variable (i.e.…”

Childhood stress and birth timing among African American women: Cortisol as biological mediator

“…Given diurnal and gestational effects on cortisol levels, hours awake prior to venipuncture and gestational age at venipuncture were also examined. An augmented backward elimination approach, which has been found to reduce bias in variable selection, was used to identify variables for the final adjusted models (Dunkler et al, 2014). Briefly, childhood stress effects were modeled using equations 1–3 above and including all variables in the working set of potential confounders.…”

Ultrasound Guidance and Other Determinants of Successful Peripheral Artery Catheterization in Critically Ill Children*

“…Covariates for multivariable models were chosen based on a purposeful selection algorithm with a significance threshold of 0.25 and a change in coefficient threshold of 20% (17). The following covariates were included in the multivariable models: patient age in years, patient weight for age, patient systolic blood pressure percentile, patient pediatric index of mortality 3 (PIM3) probability of death, fluid overload percent, and trainee months in fellowship at the time of the procedure.…”