Using decision trees to understand structure in missing data

Abstract

ObjectivesDemonstrate the application of decision trees—classification and regression trees (CARTs), and their cousins, boosted regression trees (BRTs)—to understand structure in missing data.SettingData taken from employees at 3 different industrial sites in Australia.Participants7915 observations were included.Materials and methodsThe approach was evaluated using an occupational health data set comprising results of questionnaires, medical tests and environmental monitoring. Statistical methods included standard statistical tests and the ‘rpart’ and ‘gbm’ packages for CART and BRT analyses, respectively, from the statistical software ‘R’. A simulation study was conducted to explore the capability of decision tree models in describing data with missingness artificially introduced.ResultsCART and BRT models were effective in highlighting a missingness structure in the data, related to the type of data (medical or environmental), the site in which it was collected, the number of visits, and the presence of extreme values. The simulation study revealed that CART models were able to identify variables and values responsible for inducing missingness. There was greater variation in variable importance for unstructured as compared to structured missingness.DiscussionBoth CART and BRT models were effective in describing structural missingness in data. CART models may be preferred over BRT models for exploratory analysis of missing data, and selecting variables important for predicting missingness. BRT models can show how values of other variables influence missingness, which may prove useful for researchers.ConclusionsResearchers are encouraged to use CART and BRT models to explore and understand missing data.

Physical activity patterns and associations with health-related quality of life in bladder cancer survivors

“…Item non-response was addressed first by looking for missingness patterns using aggregation plots, matrix plots, and recursive partitioning. [14, 15] Using the MICE algorithm, multiple imputation was used to generate 50 complete datasets. [15] The imputation model included key variables associated with missingness as well as variables judged useful for imputation by the MICE algorithm using influx and outflux statistics.…”

Derivation of decision rules to predict clinically important outcomes in acute flank pain patients

“…The rpart program uses a native algorithm of “surrogate splits” to handle missing data in the predictor variables (when a value for a predictor variable is missing, and that variable needs to be used to determine a split, an alternative variable that is highly correlated with the missing variable is used to determine the direction of the split) [26]. Thus, we used the entire cohort for outcomes were recorded regardless of missing data among predictors, depending on the surrogate split function.…”

Using Boosted Regression Trees and Remotely Sensed Data to Drive Decision-Making

“…BRT have also been compared favourably with other flexible regression approaches such as generalised additive models [14]. An example of BRT models helping in developing an understanding of missingness structure in the data is given by [26]. In this study…”

“…Tierney [26] concluded that more knowledge was gained about the origins of the data and the data collection process, as well as the handling of missing values for future analysis. In another study [26], the author took a different approach to deal with missing values by taking summary values such as the mean over grouped data.…”

“…In another study [26], the author took a different approach to deal with missing values by taking summary values such as the mean over grouped data.…”