Hi there,
Usually this error means one of two things:
1) There are linearly dependent or extremely highly correlated variables in
your dataset.
2) The imputation model isn't identified because you have more parameters
than observations. Note that you need far more observations than in a
typical regression because you need to run regressions for each column with
missing data.
Note that you have turned off the internal Amelia error checking, which
should give you an informative error message if there is an obvious problem
with the data. Use "incheck = TRUE" to get that. Either way, you should
probably start with a couple of variable to see if Amelia runs then and
then try to remove any highly correlated (rho > 0.95 or so) variables from
the data. You can also use the "empri" argument to stabilize the
computation. See the manual for more details on this.
Cheers,
Matt
~~~~~~~~~~~
Matthew Blackwell
Assistant Professor of Government
Harvard University
url:
http://www.mattblackwell.org
On Mon, Nov 9, 2015 at 12:31 AM W. D. <WD(a)a7h.com> wrote:
Hi,
I ran something similar to:
DataImputed <- amelia(x=GroupData,incheck=FALSE)
GroupData is several thousand observations
with about 4 dozen variables.
After iterating over a thousand times with this
error:
error: inv_sympd(): matrix appears to be singular
It ended with:
Error in emarch(x.stacked$x, p2s = p2s, thetaold = NULL,
tolerance = tolerance, :
BLAS/LAPACK routine 'DLASCL' gave error code -4
I searched online for both of these error messages
but didn't find anything useful.
Do you have any idea what might be going on, or how
to fix it?
Thanks for any light you can shed!
--
Amelia mailing list served by HUIT
[Un]Subscribe/View Archive:
http://lists.gking.harvard.edu/?info=amelia
More info about Amelia:
http://gking.harvard.edu/amelia
Amelia mailing list
Amelia(a)lists.gking.harvard.edu
To unsubscribe from this list or get other information:
https://lists.gking.harvard.edu/mailman/listinfo/amelia