Variable selection in the presence of missing data: resampling and imputation.

In the presence of missing data, variable selection methods need to be tailored to missing data mechanisms and statistical approaches used for handling missing data. We focus on the mechanism of missing at random and variable selection methods that can be combined with imputation. We investigate a general resampling approach (BI-SS) that combines bootstrap imputation and stability selection, the latter of which was developed for fully observed data. The proposed approach is general and can be applied to a wide range of settings. Our extensive simulation studies demonstrate that the performance of BI-SS is the best or close to the best and is relatively insensitive to tuning parameter values in terms of variable selection, compared with several existing methods for both low-dimensional and high-dimensional problems. The proposed approach is further illustrated using two applications, one for a low-dimensional problem and the other for a high-dimensional problem.

[1]  E. George,et al.  APPROACHES FOR BAYESIAN VARIABLE SELECTION , 1997 .

[2]  Bradley Efron,et al.  Missing Data, Imputation, and the Bootstrap , 1994 .

[3]  John Van Hoewyk,et al.  A multivariate technique for multiply imputing missing values using a sequence of regression models , 2001 .

[4]  Stef van Buuren,et al.  MICE: Multivariate Imputation by Chained Equations in R , 2011 .

[5]  R. Tibshirani,et al.  Least angle regression , 2004, math/0406456.

[6]  Francis R. Bach,et al.  Bolasso: model consistent Lasso estimation through the bootstrap , 2008, ICML '08.

[7]  Hongtu Zhu,et al.  VARIABLE SELECTION FOR REGRESSION MODELS WITH MISSING DATA. , 2010, Statistica Sinica.

[8]  Brent A. Johnson,et al.  Penalized Estimating Functions and Variable Selection in Semiparametric Regression Models , 2008, Journal of the American Statistical Association.

[9]  Qi Long,et al.  Multiple imputation in the presence of high-dimensional data , 2016, Statistical methods in medical research.

[10]  Nicole A. Lazar,et al.  Statistical Analysis With Missing Data , 2003, Technometrics.

[11]  Willem van Mechelen,et al.  Variable selection under multiple imputation using the bootstrap in a prognostic study , 2007, BMC medical research methodology.

[12]  Francis R. Bach,et al.  Model-Consistent Sparse Estimation through the Bootstrap , 2009, ArXiv.

[13]  J. Ibrahim,et al.  Variable Selection in the Cox Regression Model with Covariates Missing at Random , 2010, Biometrics.

[14]  Patrick Royston,et al.  How should variable selection be performed with multiply imputed data? , 2008, Statistics in medicine.

[15]  S. Lahiri,et al.  Bootstrapping Lasso Estimators , 2011 .

[16]  R. Tibshirani.  Regression Shrinkage and Selection via the Lasso , 1996 .

[17]  Thomas R Belin,et al.  Imputation and Variable Selection in Linear Regression Models with Missing Covariates , 2005, Biometrics.

[18]  Sijian Wang,et al.  Variable Selection for Multiply-imputed Data with Application to Dioxin Exposure Study Variable Selection for Multiply-imputed Data , 2011 .

[19]  N. Meinshausen,et al.  Stability selection , 2008, 0809.2932.

[20]  Variable selection when missing values are present: a case study , 2011, Statistical methods in medical research.

[21]  Tso-Jung Yen,et al.  Discussion on "Stability Selection" by Meinshausen and Buhlmann , 2010 .

[22]  H. Akaike.  A new look at the statistical model identification , 1974 .

[23]  A. Gelman,et al.  Multiple Imputation with Diagnostics (mi) in R: Opening Windows into the Black Box , 2011 .

[24]  J. Wolfson.  EEBoost: A General Method for Prediction and Variable Selection Based on Estimating Equations , 2011 .

[25]  A. Tsiatis.  Semiparametric Theory and Missing Data , 2006 .