[R] Changing MaxNWts with the mi() function (error message)

[Heather Kettrey]

Hello,

I am trying to impute data with the mi() function (mi package) and
keep receiving an error message. When imputing the variable, "sex,"
the mi() function accesses the mi.categorical() function, which then
accesses the nnet() function. I then receive the following error
message (preceded by my code below):


> imputed.england=mi(england.pre.imputed, n.iter=6, add.noise=FALSE)
Beginning Multiple Imputation ( Mon Jan 14 13:39:49 2013 ):
Iteration 1
 Chain 1 : sex
Error while imputing variable: sex , model: mi.categorical
 Error in nnet.default(X, Y, w, mask = mask, size = 0, skip = TRUE,
softmax = TRUE,  :
  too many (3432) weights


The error message indicates that there are too many weights (3432). I
know you can adjust the maximum weights with the mi.categorical()
function (i.e., MaxNWts= 3500), but I do not see a way to do this with
the mi() function. I found an R-help posting from two years ago with
some instruction on how to change the max weights in mi(), but I can't
seem to get it to work. The following is a link to the post:

http://r.789695.n4.nabble.com/Changing-model-parameters-in-the-mi-package-td2288851.html#a2288933

And my implementation of the suggestion from that post:

> info=mi.info(england.pre.imputed)
> info[["sex"]]$params$MaxNWts=3500
> info[["sex"]]$params
$n.iter
[1] 100

$MaxNWts
[1] 3500

$missing.index
NULL

When I try the imputation again I receive the same error message as before:

> imputed.england=mi(england.pre.imputed, n.iter=6, add.noise=FALSE)
Beginning Multiple Imputation ( Mon Jan 14 14:20:46 2013 ):
Iteration 1
 Chain 1 : sex*
Error while imputing variable: sex , model: mi.categorical
 Error in nnet.default(X, Y, w, mask = mask, size = 0, skip = TRUE,
softmax = TRUE,  :
  too many (3432) weights

So, I tried a few more things, and received the same error message.

First...............
> imputed.england=mi(england.pre.imputed,info=info)
This yields the same error message.

Then...............
> inf=mi.info(england.pre.imputed)
> inf=update(inf, "MaxNWts", list(sex=3500))
> imputed.england=mi(england.pre.imputed, info=inf)
Beginning Multiple Imputation ( Mon Jan 14 14:31:20 2013 ):
Iteration 1
 Chain 1 : sex*
Error while imputing variable: sex , model: mi.categorical
 Error in nnet.default(X, Y, w, mask = mask, size = 0, skip = TRUE,
softmax = TRUE,  :
  too many (3432) weights

Again, the same error message.

For some reason, I am unable to successfully adjust the max weights. I
would greatly appreciate any suggestions.

Thanks in advance!

-- 
Heather Hensman Kettrey
PhD Candidate
Department of Sociology
Vanderbilt University