Abstract:
|
Missing data are a commonly occurring phenomenon. Advances in computational statistics have produced flexible missing-data procedures with a sound statistical basis. One of these procedures involves multiple imputation (MI), which is a stochastic simulation technique in which the missing values are replaced by simulated versions. Most incomplete data sets involve variables of many different types on a structural level; interdependencies are a function of a mixture of binary, ordinal, count, and continuous variables as well as nonresponse rates and mechanisms, all of which act simultaneously to characterize the data-analytics paradigms under consideration. The use of digital data is growing rapidly as researchers get more capable of collecting instantaneous self-reported data . In this context, planned missing data designs have been developed to reduce respondent burden and lower the cost associated with data collection. This talk is concerned with theoretical, algorithmic, and implementation-based components of a unified, general-purpose Bayesian multiple imputation framework for intensive data sets that are collected via increasingly popular real-time data capture approaches.
|