TY - GEN

T1 - Generalizing Univariate Predictive Mean Matching to Impute Multiple Variables Simultaneously

AU - Cai, Mingyang

AU - van Buuren, Stef

AU - Vink, Gerko

N1 - Publisher Copyright:
© 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.

PY - 2022/7/7

Y1 - 2022/7/7

N2 - Predictive mean matching (PMM) is an easy-to-use and versatile univariate imputation approach. It is robust against transformations of the incomplete variable and violation of the normal model. However, univariate imputation methods cannot directly preserve multivariate relations in the imputed data. We wish to extend PMM to a multivariate method to produce imputations that are consistent with the knowledge of derived data (e.g., data transformations, interactions, sum restrictions, range restrictions, and polynomials). This paper proposes multivariate predictive mean matching (MPMM), which can impute incomplete variables simultaneously. Instead of the normal linear model, we apply canonical regression analysis to calculate the predicted value used for donor selection. To evaluate the performance of MPMM, we compared it with other imputation approaches under four scenarios: 1) multivariate normal distributed data, 2) linear regression with quadratic terms; 3) linear regression with interaction terms; 4) incomplete data with inequality restrictions. The simulation study shows that with moderate missingness patterns, MPMM provides plausible imputations at the univariate level and preserves relations in the data.

AB - Predictive mean matching (PMM) is an easy-to-use and versatile univariate imputation approach. It is robust against transformations of the incomplete variable and violation of the normal model. However, univariate imputation methods cannot directly preserve multivariate relations in the imputed data. We wish to extend PMM to a multivariate method to produce imputations that are consistent with the knowledge of derived data (e.g., data transformations, interactions, sum restrictions, range restrictions, and polynomials). This paper proposes multivariate predictive mean matching (MPMM), which can impute incomplete variables simultaneously. Instead of the normal linear model, we apply canonical regression analysis to calculate the predicted value used for donor selection. To evaluate the performance of MPMM, we compared it with other imputation approaches under four scenarios: 1) multivariate normal distributed data, 2) linear regression with quadratic terms; 3) linear regression with interaction terms; 4) incomplete data with inequality restrictions. The simulation study shows that with moderate missingness patterns, MPMM provides plausible imputations at the univariate level and preserves relations in the data.

KW - Block imputation

KW - Canonical regression analysis

KW - Missing data

KW - Multiple imputation

KW - Multivariate analysis

KW - Predictive mean matching

UR - http://www.scopus.com/inward/record.url?scp=85135097397&partnerID=8YFLogxK

U2 - 10.1007/978-3-031-10461-9_5

DO - 10.1007/978-3-031-10461-9_5

M3 - Conference contribution

AN - SCOPUS:85135097397

SN - 978-3-031-10460-2

T3 - Lecture Notes in Networks and Systems

SP - 75

EP - 91

BT - Intelligent Computing

A2 - Arai, Kohei

PB - Springer

CY - Cham

T2 - Computing Conference, 2022

Y2 - 14 July 2022 through 15 July 2022

ER -