I'm trying to analyze complex survey data with Survey. I imputed missing data with mice and, following the instructions in the documentation, have converted the imputations to an imputationList object with imputationList()in mitools. But when I try to use that object as data in svydesign(), I get this error message:
Error in as.data.frame.default(yrbs_complex_imputationList) :
cannot coerce class ‘"imputationList"’ to a data.frame
Following an example provided elsewhere in StackOverflow, I tried to incorporate the mitools function directly into the svydesign formula:
yrbs_svyimputationList<-svydesign(ids="psu", probs = NULL, strata = "stratum", variables = NULL, fpc = NULL, data=imputationList(yrbs_complex_imputations), nest = TRUE, check.strata = !nest, weights, pps=FALSE)
But this resulted in a different error message:
Error in as.data.frame.default(x[[i]], optional = TRUE) :
cannot coerce class ‘"function"’ to a data.frame
How can I incorporate the multiple-imputed data into a survey design object?
Here's the example from the documentation
> library(mitools)
> data.dir<-system.file("dta",package="mitools")
> files.men<-list.files(data.dir,pattern="m.\\.dta$",full=TRUE)
> men<-imputationList(lapply(files.men, foreign::read.dta,
+ warn.missing.labels=FALSE))
> files.women<-list.files(data.dir,pattern="f.\\.dta$",full=TRUE)
> women<-imputationList(lapply(files.women, foreign::read.dta,
+ warn.missing.labels=FALSE))
> men<-update(men, sex=1)
> women<-update(women,sex=0)
> all<-rbind(men,women)
>
> designs<-svydesign(id=~id, strata=~sex, data=all)
> designs
Multiple (5) imputations: svydesign(id = ~id, strata = ~sex, data = all)
The big difference is that you need to use ~ rather than quotation marks to quote variables, just as in regression models. Nowadays this might get implemented using the non-standard evaluation from the tidyverse, but the survey package is considerably older than the tidyverse.