On the use of multiple imputation to address data missing by design as well as unintended missing data in case-cohort studies with a binary endpoint

Table 2 Summary of analysis approaches used in the simulation and case studies

Approach	Imputation Approach	Weights in analysis model	How missing data addressed		Imputation sample	Analysis sample
Approach	Imputation Approach	Weights in analysis model	‘by design’	‘by chance’	Imputation sample	Analysis sample
CCA	None	sampling	IPW	Not addressed	N/A	Subset
IPW-Only	None	combined^a	IPW	IPW	N/A	Subset
MI-IPW-Sub	FCS with outcome only	sampling	IPW	MI	Subset	Subset
MI-IPW-Int	FCS with outcome only	sampling	IPW	MI	Full cohort	Subset
MI-Only	FCS imputation	None	MI	MI	Full cohort	Full cohort

FCS Fully conditional specification, IPW Inverse probability weighting, MI Multiple imputation
^aCombined weights are the product of inverse probability of selection into subcohort and inverse probability of being a complete observation

ISSN: 1471-2288