Multiple Imputation for Multivariate Data with Missing and Below‐Threshold Measurements: Time‐Series Concentrations of Pollutants in the Arctic

Abstract

Many chemical and environmental data sets are complicated by the existence of fully missing values or censored values known to lie below detection thresholds. For example, week-long samples of airborne particulate matter were obtained at Alert, NWT, Canada, between 1980 and 1991, where some of the concentrations of 24 particulate constituents were coarsened in the sense of being either fully missing or below detection limits. To facilitate scientific analysis, it is appealing to create complete data by filling in missing values so that standard complete-data methods can be applied. We briefly review commonly used strategies for handling missing values and focus on the multiple-imputation approach, which generally leads to valid inferences when faced with missing data. Three statistical models are developed for multiply imputing the missing values of airborne particulate matter. We expect that these models are useful for creating multiple imputations in a variety of incomplete multivariate time series data sets.

Incidence rates of erectile dysfunction in the Dutch general population. Effects of definition, clinical relevance and duration of follow-up in the Krimpen Study

“…It has been discussed extensively that the complete case analysis is an incomplete analysis when missing data occur. [13][14][15][16][17][18][19][20] Additionally, we analysed all data using IVEware for a multiple imputation strategy (MI). 13,[15][16][17]19 To impute the missing ED status IVEware used a model with all baseline data and the known ED status in first and second follow-up.…”

Anthropogenic and natural constituents in particulate matter in the Netherlands

“…To independently address such issues one can apply statistical techniques to the observation dataset to address the possible source categories for PM. Techniques such as positive matrix factorization (PMF) can be used to search for groups of components with the same behaviour in time, which are then assumed to derive from a single source or group of similar sources (Hopke et al, 2001). Hence, a mass closure approach as well as PMF will give us independent insights from an empirical point of view in the actual PM sources.…”

Unintended Consequences of a Standard Admission Order Set on Venous Thromboembolism Prophylaxis and Patient Outcomes

“…Since UCSF only began recording present on admission designations in January 2007, we used multiple imputation to predict present on admission designations for patients admitted to the hospital prior to this. [17][18][19] Using a stepwise procedure, we developed logistic models for present on admission designations-either "yes" or any other nonmissing option (coded "no" to maximize positive predictive value of the "yes" designation)-for VTE, intracranial hemorrhage, gastrointestinal hemorrhage, other hemorrhage, liver disease, coagulopathy, and sepsis (the only contributor to the ACCP criteria for "likely benefit" that could plausibly develop in the hospital), based on data obtained in January 2007 and later. Each model contained different predictors to better approximate the diverse outcomes noted above, but in general included demographic and clinical variables such as sex, race, and variable Elixhauser comorbidity measures.…”

“…In all analyses using these five completed data sets, we used standard methods to combine results, ensuring that overall inferences properly reflected the additional uncertainty induced by the missing present on admission designations. [17][18][19][20] To assess the accuracy of our imputation approach, we reviewed the 226 charts above, which were all missing the VTE present on admission designations. The expected present on admission designations by imputation were extremely well calibrated to observed presence on admission designation by chart review (HosmerLemeshow test, p=0.94).…”