[R] Panel data and imputed datasets
napaxton at fas.harvard.edu
Tue Aug 14 23:50:24 CEST 2007
I am hardly an expert, so I expect that this code is not the easiest/
most efficient way of getting where I want. Any suggestions in that
direction would also be helpful.
I am working on panel analysis with five imputed datasets, generated
by Amelia. To do panel analysis, it seemed that the plm package was
the best, providing a convenient wrapper for fixed and random effects
and Hausman tests. And since I've got five datasets, it seemed
reasonable that what I wanted to do was to combine these as a big
imputation data frame. The plm regression function requires a data
argument, and simply using the imputationList result didn't work, nor
did calling the names(allhiv) work (that resulted in "imputations"
Where do I point the plm command to look for the data? Or have I
gone about this in entirely the wrong way?
Thanks in advance for the help.
Here's the code I have so far:
#Read in the imputed datasets
data1 <- read.dta("outdata1.dta")
data2 <- read.dta("outdata2.dta")
data3 <- read.dta("outdata3.dta")
data4 <- read.dta("outdata4.dta")
data5 <- read.dta("outdata5.dta")
#Set the datasets as panels (for subsequent analysis in plm)
pdata.frame(data1, "country", time="year")
pdata.frame(data2, "country", time="year")
pdata.frame(data3, "country", time="year")
pdata.frame(data4, "country", time="year")
pdata.frame(data5, "country", time="year")
#Using mitools, combine the panel data frames 1-5
allhiv <- imputationList(list(data1,data2,data3,data4,data5))
#A test regression
#The following is something I tried but which didn't work.
estimateHIV1 <-with(allhiv, plm(form2, data=imputations[1:5,]))
Nathan A. Paxton
Dept. of Government, Harvard University
John Winthrop House, Harvard University
napaxton AT fas DOT harvard DOT edu
When you have to stay eight years away from California, you live in a
perpetual state of homesickness.
- Ronald Reagan
The most courageous act is still to think for yourself. Aloud.
More information about the R-help