Missing Data Project

Topic Leader Status Most Recent
Updated Time
Compare MICE, argeimpute, and/or probably some maximum
likelihood methods for normal distribution up to 99% missing
Illustration of situations where prediction mean matching may
fail,and to develop a diagnostic how a researcher can detect
a priori that he/she is in such situation
Study the situation when there are multiple covariates and/or
outcome missing simultaneously, and check how bad it can get
Study categorical missing covariates using Fisher's optimum
scoring algorithm, and study how good the prediction mean
match based on canonical variable scores is comparing to
polytomous logistic full model or MICE
Over-fitting (or over-imputing) issue      
Compare CC, MICE, and dropping missing variable for diagnostic
data. There will be one continuous variable and two binary variables
in two settings: (a) continous variable is missing (b)one of the binary
variables is missing The missing percentage varies from 10% to 90%.
Kristel working Nov 27,2006
Compare several methods to impute a missing value when a prediction rule is implemented in daily clinical practice. Two predictions rules are used: a diagnostic simplified score and a prognostic rule. The imputation methods include imputation of the mean (or median), the subgroup mean (or median), the use of the rule without the predictor with the missing values (by using the one-step-sweep method) and out of sample multiple imputation (the derivation data is stored and used to multiple impute the missing value). Kristel working Dec 22,2006

* Set ALLOWTOPICCHANGE = BiostatUsersGroup
Topic revision: r6 - 22 Dec 2006, KristelJanssen

This site is powered by FoswikiCopyright © 2013-2022 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding Vanderbilt Biostatistics Wiki? Send feedback