Multiple Imputation for Combined-Survey Estimation With Incomplete Regressors In One But Not Both Surveys

Abstract

This product is part of the RAND Labor and Population working paper series. RAND working papers are intended to share researchers' latest findings and to solicit informal peer review. They have been approved for circulation by RAND Labor and Population but have not been formally edited or peer reviewed. Unless otherwise indicated, working papers can be quoted and cited without permission of the author, provided the source is clearly referred to as a working paper. RAND's publications do not necessarily reflect the opinions of its research clients and sponsors. is a registered trademark.

Associations between low- and high-dose oral fluconazole and pregnancy outcomes: 3 nested case–control studies

“…Where data were missing for any of the sociodemographic variables, we used the method of monotone multiple imputation. 28 We used SAS for all statistical analyses (SAS Institute Inc., version 9.2). Differences were significant when the 95% CIs did not overlap 1.0 and when p < 0.05 (2-tailed).…”

“…23 Furthermore, linkage between maternal and child data in the Quebec Pregnancy Cohort was possible more than 95% of the time. [24][25][26][27] Once a woman and her child are included in a study, we have access to all variables (available data); when data were missing for any of the sociodemographic variables, we used the accepted method of monotone multiple imputation 28 (Supplementary Figure S2, Appendix 1).…”

“Discrimination” versus “Unfair Treatment”: Measuring Differential Treatment and its Association with Health

“…By accounting for the effects of sociodemographic characteristics, the remaining differences across surveys can be interpreted as, at least partially, the effect of question wording. Prior research demonstrates that Chow test can be used to estimate whether the effect of a variable on an outcome varies across groups (Long and Mustillo forthcoming; also see Rendall et al 2013 for a review). Table 1 presents the descriptive statistics for all variables included in the current analyses in both MIDUS ("discrimination") and HRS ("unfair treatment") surveys.…”

Cross-country cross-survey design in international marketing research

“…According to Rendall et al (2013), cross-survey missingness is monotonic and easily satisfies the MAR assumption needed for unbiased multiple imputation. For example, if two surveys, Survey 1 with fewer variables (X1, X2 and X3) and Survey 2 with more variables (X1, X2, X3, X4, and X5) are collected separately and combined.…”

“…Specifically, in international marketing research, it is assumed that in cross-country research, country-specific samples represent the same population and are thus drawn from the common universe (Rendall et al, 2013). This assumption needs to be carefully studied prior to data collection.…”