A Potential for Bias When Rounding in Multiple Imputation

Horton, Nicholas J.; Lipsitz, Stuart R.; Parzen, Michael

doi:10.1198/0003130032314

Cited by 181 publications

(155 citation statements)

References 2 publications

Supporting

Mentioning

148

Contrasting

Unclassified

Order By: Relevance

“…Although continuous variables can be imputed with unbiased results, the same cannot be said of binary variables (such as the remission measure which we imputed) (Horton et al, 2003). Because of this issue of bias, and because we found the same pattern of results in the models using imputed data as those using list-wise deletion, we present only the latter models.…”

Section: Analysesmentioning

confidence: 98%

The role of medical conditions and primary care services in 5-year substance use outcomes among chemical dependency treatment patients

Mertens¹,

Flisher²,

Satre³

et al. 2008

Drug and Alcohol Dependence

View full text Add to dashboard Cite

Introduction-Health problems are prevalent in chemical dependency (CD) treatment populations, and often precede reductions in substance use among untreated populations. Few studies have examined whether medical problems predict better long-term outcomes in treated individuals, or how primary care utilization and CD/primary care service integration affects long-term outcomes among those with health problems Method-In a sample of 598 CD patients in a private health plan, logistic regression models examined whether Substance Abuse-related Medical Conditions (SAMCs), integrated medical and CD care, and on-going primary care predicted remission of CD problems at 5 years.Results-Those with SAMCs were no more likely than others to be remitted at 5 years except among young adults and those with medical, but not psychiatric SAMCs. Higher levels of medical problem severity at intake and receiving integrated CD and primary care in the index treatment episode predicted remission in the full sample and among those with SAMCs. Among those with SAMCs, individuals with ongoing medical care-two to ten primary care visits-in the 5 years following intake were more likely to be remitted at 5 years than those with fewer visits.Conclusions-This study highlights the potentially important role of medical services in the longterm treatment of CD disorders. CD treatment may benefit from a disease management approach similar to that recommended for other chronic medical problems: specialty care when the condition is severe followed by services in primary care when the condition is stabilized.

show abstract

Section: Analysesmentioning

confidence: 98%

The role of medical conditions and primary care services in 5-year substance use outcomes among chemical dependency treatment patients

Mertens¹,

Flisher²,

Satre³

et al. 2008

Drug and Alcohol Dependence

View full text Add to dashboard Cite

show abstract

“…The resulting imputed values will look implausible if inspected closely, but that often has little effect on the analytic results. In fact, attempts to edit the imputed values to improve their plausibility can introduce bias by changing the variables' means, variance, and covariances (Horton et al 2003;Allison 2005;Bernaards et al 2007;von Hippel 2009). …”

Section: Introductionmentioning

confidence: 99%

“…For example, in some settings it can be acceptable to impute dummy variables, squared terms, and interactions as though they were normal when conditioned on other variables (Horton, Lipsitz, and Parzen 2003;Allison 2005;Bernaards, Belin, and Schafer 2007;von Hippel 2009). The resulting imputed values will look implausible if inspected closely, but that often has little effect on the analytic results.…”

Section: Introductionmentioning

confidence: 99%

Should a Normal Imputation Model be Modified to Impute Skewed Variables?

Hippel

2012

Sociological Methods & Research

View full text Add to dashboard Cite

(169 words)Researchers often impute continuous variables under an assumption of normality, yet many incomplete variables are skewed. We find that imputing skewed continuous variables under a normal model can lead to bias; the bias is usually mild for popular estimands such as means, standard deviations, and linear regression coefficients, but the bias can be severe for more shapedependent estimands such as percentiles or the coefficient of skewness. We test several methods for adapting a normal imputation model to accommodate skewness, including methods that transform, truncate, or censor (round) normally imputed values, as well as methods that impute values from a quadratic or truncated regression. None of these modifications reliably reduces the biases of the normal model, and some modifications can make the biases much worse. We conclude that, if one has to impute a skewed variable under a normal model, it is usually safest to do so without modifications-unless you are more interested in estimating percentiles and shape that in estimated means, variance, and regressions. In the conclusion, we briefly discuss promising developments in the area of continuous imputation models that do not assume normality.

show abstract

“…Recent research, however, has shown that these modifications tend to cause more problems than they solve. Rounding off imputed values for categorical variables has been found to produce biased parameter estimates because such values are typically not normally distributed around the cutoff point (for instance, 0.5 in the case of binary variables) (Horton, Lipsitz, and Parzen 2003;Allison 2005;Cranmer and Gill 2013). Transforming skewed variables has also been found to increase bias because it alters their relationship with other variables in the imputation model; in effect, it is equivalent to assuming that they have zero correlation with such variables (von Hippel 2013).…”

mentioning

confidence: 99%

How Multiple Imputation Makes a Difference

Lall

2016

Polit. anal.

174

View full text Add to dashboard Cite

Edited by R. Michael AlvarezPolitical scientists increasingly recognize that multiple imputation represents a superior strategy for analyzing missing data to the widely used method of listwise deletion. However, there has been little systematic investigation of how multiple imputation affects existing empirical knowledge in the discipline. This article presents the first large-scale examination of the empirical effects of substituting multiple imputation for listwise deletion in political science. The examination focuses on research in the major subfield of comparative and international political economy (CIPE) as an illustrative example. Specifically, I use multiple imputation to reanalyze the results of almost every quantitative CIPE study published during a recent fiveyear period in International Organization and World Politics, two of the leading subfield journals in CIPE. The outcome is striking: in almost half of the studies, key results "disappear" (by conventional statistical standards) when reanalyzed.

show abstract

A Potential for Bias When Rounding in Multiple Imputation

Cited by 181 publications

References 2 publications

The role of medical conditions and primary care services in 5-year substance use outcomes among chemical dependency treatment patients

The role of medical conditions and primary care services in 5-year substance use outcomes among chemical dependency treatment patients

Should a Normal Imputation Model be Modified to Impute Skewed Variables?

How Multiple Imputation Makes a Difference

Contact Info

Product

Resources

About