Stepwise Regression in Social and Psychological Research

Abstract

Researchers often invoke stepwise ordinary least squares regression to explain, predict or classify practical problems or theoretical constructs in psychological and social research. Unfortunately, this statistical technique is used without proper consideration for its inherent theoretical and practical limitations, a problem expected to grow even more serious with the proliferation of statistical packages for use on personal computers. Use of stepwise regression in social and psychological research is reconsidered here. Explanations of forward selection, backward elimination and combination stepwise procedures are provided; limitations of the technique, statistical and practical, are then addressed. Analysis shows that most of the current applications of stepwise regression should be rejected, or at least tempered with strong qualification to inference.

Significant Predictors of Soccer Players’ Moral Functioning from Components of Contextual Injustice, Sensitivity to Injustice and Moral Atmosphere

“…Given that this study is the first to evaluate and characterise the simultaneous influence of these variables, we also did not want to introduce the variables in a particular order as no theoretical element justifies the higher importance of any of these variables. The assertions from Huberty (1989) and Henderson and Denison (1989) concerning the danger of the step-by-step method, which can introduce bias in the analysis, supported our decision. Table 1 presents the means and standard deviations for each of the dependent variables and predictors according to the condition.…”

The Relationship Between Time and Birth Mother Satisfaction With Relinquishment

“…Given the lack of prior research and theory available to guide the variable entry selection process, a forward stepwise approach was selected for this analysis. Stepwise regression is an appropriate technique for evaluating the importance of variables and for selecting the most useful subsets of predictor variables (Henderson & Denison, 1989). The strongest predictor was included in the first model, and the strongest predictor of the remaining variables was added to subsequent models (p < .05, and lowest p value).…”

Characteristics and role in outcome prediction of continuous EEG after status epilepticus: A prospective observational cohort

“…Still, these particular terms cannot be applied to all RPPs, and the relatively poor interrater agreement reported 8 may lessen the predictive value of the “plus modifier.” Furthermore, stepwise backward logistic regression model was employed in the present study to identify variables with independent prediction value, a technique that some consider suboptimal. 32,33 However, a stepwise regression model in logistic regression is an efficient and widely used way to analyze the effect of a group of independent variables on a binary outcome. 34,35 Careful variable selection was made (know predictors, statistically significant variables in the univariate analysis) and necessary assumptions for regression 34 were checked (absence of colinearity, numbers of variable) to minimize the potential limitations of this procedure.…”