New subject: Amelia Error

25 Oct 2006

I just set them to the ranges of the observed variables.  I didn't want the
imputed values to lie very far outside the observed range. 

-----Original Message-----
From: Matthew Blackwell [mailto:blackwel@fas.harvard.edu] 
Sent: Wednesday, October 25, 2006 1:44 PM
To: david.armstrong(a)politics.ox.ac.uk
Cc: amelia(a)lists.gking.harvard.edu
Subject: Re: [amelia] Amelia Error 

Hello Dave,

The reason you are getting errors on 'mu.priors' and not the ranges is 
that the ranges are simply converted to a normal distribution with a mean 
and standard deviation that match the ranges given.  So, my guess is that 
something strange is happening with your range priors settings.

Would you mind telling us what your settings for mins, maxs and conf are?

Thanks,
matt.

On Wed, 25 Oct 2006, Gary King wrote:

...
  here's a message that just arrived.   are you on
the listserv?
 can you respond?
 thanks,
 Gary

 ---------- Forwarded message ----------
 Date: Wed, 25 Oct 2006 11:48:17 +0100
 From: Dave Armstrong &lt;david.armstrong(a)politics.ox.ac.uk&gt;
 To:  &lt;amelia(a)lists.gking.harvard.edu&gt;
 Subject: [amelia] Amelia Error

 Dear Amelia Users,

 This may be a problem with my understanding of how Amelia is supposed to
 work, but I've encountered the following problem.  I give Amelia the
 following command:

 amelia.out <- amelia(smalldat, cs=1, ts=2, polytime=3, lags=1, leads=1,
 intercs=F, incheck=F, mins=mins, maxs=maxs, conf=conf, means=NULL, sds=NULL,
...
  logs=c(5,6,7,8),

 sqrts=c(9,10,11,12,13,14,15,16),
 noms=c(17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32))

 and I get the following error:

 Error in mu.priors - (ones %*% meanx) : non-conformable arrays

 using traceback() I tracked it down to the scalecenter subcommand.  Since
 I've not provided any means and standard deviations for the priors, rather
 I've provided
 ranges, I assume that setting means and sds both to NULL will suggest that
 mu.priors and sd.priors will also be NULL.  If I comment out the troublesome
...
  piece of
 the scalecenter command as below, source this command in and issue the same
...
  command as above to Amelia, it works fine.

 scalecenter <- function (x, mu.priors = NULL, sd.priors = NULL)
 {
    AMn <- nrow(x)
    ones <- matrix(1, AMn, 1)
    meanx <- colMeans(x, na.rm = TRUE)
    stdvx <- apply(x, 2, sd, na.rm = TRUE)
    x.ztrans <- (x - (ones %*% meanx))/(ones %*% stdvx)
 #    if (!identical(mu.priors, NULL)) {
 #        mu.priors <- (mu.priors - (ones %*% meanx))/(ones %*%
 #            stdvx)
 #        sd.priors <- sd.priors/(ones %*% stdvx)
 #    }
    return(list(x = x.ztrans, mu = meanx, sd = stdvx, mu.priors = NULL,
        sd.priors = NULL))
 }

 Can you give me some advice as to what I'm doing wrong in the first step?

 Thanks,
 Dave.

 Dave Armstrong
 Department of Politics and International Relations
 Manor Road
 Oxford
 OX1 3UQ
 Phone: 08165 285956
 E-mail: david.armstrong(a)politics.ox.ac.uk
 Web: http://users.ox.ac.uk/~polf0104

-
Amelia mailing list served by Harvard-MIT Data Center
[Un]Subscribe/View Archive: http://lists.gking.harvard.edu/?info=amelia

RE: [amelia] Amelia Error