[R] manipulating multiply imputed data sets

[John Fox]

Dear Donald,

I can't guarantee that there aren't other problems, but your call to
update() is in error; you need 

allmis <- update(allmis, white = recode(RACE, '1=1; else=0; '))

not

allmis <- update(allmis, white <- recode(RACE, '1=1; else=0; '))

[The last ; in the recode specification is unnecessary, but should do no
harm; as well, for a simple recode like this you might prefer ifelse() to
recode().]

Here's an example:

--------- snip ---------

> library(mitools)
> library(car)
> data(smi)
> smi <- update(smi, sex=recode(sex, "0 = 'M'; 1 = 'F'"))
> with(smi, table(sex, drkfre))
[[1]]
   drkfre
sex Non drinker not in last wk <3 days last wk >=3 days last wk
  F         207            194             134               35
  M         282            201             105               12

[[2]]
   drkfre
sex Non drinker not in last wk <3 days last wk >=3 days last wk
  F         200            200             132               38
  M         282            195             109               14

. . .

attr(,"call")
with.imputationList(smi, table(sex, drkfre))

----------- snip -----------

I hope this helps,
 John

------------------------------
John Fox, Professor
Department of Sociology
McMaster University
Hamilton, Ontario, Canada
web: socserv.mcmaster.ca/jfox

> -----Original Message-----
> From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org]
On
> Behalf Of Donald Braman
> Sent: May-28-08 2:21 PM
> To: r-help at r-project.org
> Subject: [R] manipulating multiply imputed data sets
> 
> Hi folks,
> 
> I have five imputed data sets and would like to apply the same
> recoding routines to each.  I could do this sort of thing pretty
> easily in Stata using MIM, but I've decided to go cold turkey on other
> stats packages as a incentive for learning more about R.  Most of the
> recoding is for nominal variables, like race, religion, urbanicity,
> and the like.  So, for example, to recode race for my first dataset,
> inmi1, I would do the following:
> 
> miset1$white  <- recode(miset1$RACE, '1=1; else=0; ')
> miset1$black  <- recode(miset1$RACE, '2=1; else=0; ')
> miset1$asian  <- recode(miset1$RACE, '3=1; else=0; ')
> miset1$hispanic <- recode(miset1$RACE, '4=1; else=0; ')
> miset1$raceother <- recode(miset1$RACE, '5=1; else=0; ')
> 
> I've tried a number of variations, e.g., on the following using recode
> (from the car package) with imputationList (from the mitools package),
> though without success:
> 
> files.allmisets <- list.files(getwd(),pattern="miset*.csv$",full=TRUE)
> allmis <- imputationList(lapply(files.allmisets, read.csv))
> allmis <- update(allmis, white <- recode(RACE, '1=1; else=0; '))
> 
> I've also tried some basic loops.  I guess I'm also a bit confused as
> to when R references the original object and when it creates a new
> one. I suppose I could do this in Python and the use PyR, but I'd
> really like to learn a bit more about how R syntax.
> 
> Any help on this specific problem or general advice on manipulating
> data in multiply imputed datasets in R would be much appreciated.
> 
> --
> Donald Braman
> http://www.law.gwu.edu/Faculty/profile.aspx?id=10123
> http://research.yale.edu/culturalcognition
> http://ssrn.com/author=286206
> 
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.