Comparison of methods for imputing limited-range variables: a simulation study

Abstract

BackgroundMultiple imputation (MI) was developed as a method to enable valid inferences to be obtained in the presence of missing data rather than to re-create the missing values. Within the applied setting, it remains unclear how important it is that imputed values should be plausible for individual observations. One variable type for which MI may lead to implausible values is a limited-range variable, where imputed values may fall outside the observable range. The aim of this work was to compare methods for imputing limited-range variables, with a focus on those that restrict the range of the imputed values.MethodsUsing data from a study of adolescent health, we consider three variables based on responses to the General Health Questionnaire (GHQ), a tool for detecting minor psychiatric illness. These variables, based on different scoring methods for the GHQ, resulted in three continuous distributions with mild, moderate and severe positive skewness. In an otherwise complete dataset, we set 33% of the GHQ observations to missing completely at random or missing at random; repeating this process to create 1000 datasets with incomplete data for each scenario.For each dataset, we imputed values on the raw scale and following a zero-skewness log transformation using: univariate regression with no rounding; post-imputation rounding; truncated normal regression; and predictive mean matching. We estimated the marginal mean of the GHQ and the association between the GHQ and a fully observed binary outcome, comparing the results with complete data statistics.ResultsImputation with no rounding performed well when applied to data on the raw scale. Post-imputation rounding and imputation using truncated normal regression produced higher marginal means than the complete data estimate when data had a moderate or severe skew, and this was associated with under-coverage of the complete data estimate. Predictive mean matching also produced under-coverage of the complete data estimate. For the estimate of association, all methods produced similar estimates to the complete data.ConclusionsFor data with a limited range, multiple imputation using techniques that restrict the range of imputed values can result in biased estimates for the marginal mean when data are highly skewed.

Sustained Posttransplantation Diabetes Is Associated With Long-Term Major Cardiovascular Events Following Liver Transplantation

“…This semi-parametric matching method uses only observed values to generate a linear prediction model, such that the distribution and range of the data are preserved and plausible imputed values are guaranteed. (31)…”

A Pilot, Multicentre Pragmatic Randomised Trial to Explore the Impact of Carer Skills Training on Carer and Patient Behaviours: Testing the Cognitive Interpersonal Model in Adolescent Anorexia Nervosa

“…As some patients only had one carer, imputation for ‘carer 2’ was conditional on their existence. One hundred imputations were generated and combined in the analysis models using Rubin's rules (Rodwell, Lee, Romaniuk, & Carlin, ). For a limited range of variables, we allowed values to be imputed outside bounds to ensure correct coverage of confidence intervals.…”

Faculty Motivation to Mentor Students Through Undergraduate Research Programs: A Study of Enabling and Constraining Factors

“…We analyzed all independent variables based on ordinal measures (derived from survey responses to Likert-type scales) as continuous predictors. This approach is considered a best practice in MI when imputing missing data and estimating model parameters, since rounding off imputed values based on discrete categorical specifications has been shown to produce more biased parameter estimates in analysis models (Allison 2005; Enders 2010; Horton et al 2003; Rodwell et al 2014). Before analyzing the data, we excluded cases with a relatively high proportion of missing data (i.e., more than 50 % missing for the variables included in the analysis), which resulted in the loss of 22 cases.…”