Hi Amanda,
Most likely there is an issue in the data that is causing the
numerical instability. First, I would try to run Amelia without the
time-trends, to see how it performs. In situations with many units and
few time periods, adding a time trend for each unit can vastly
increase the number of parameters to be estimated.
Second, you may have variables that are linear combinations of others
in your data. If I am interpreting your input correctly, this might be
the problem since
log(GDP per capita) = log(GDP) - log(population),
and each of those appear to be in the imputation model. You might want
to exclude one of the GDP measures to see if that improves the
situation.
If neither of those things helps your situation, you may found a bug.
In that case, please contact me off-list and we can troubleshoot the
error. Hope that helps.
Cheers,
matt.
On Wed, Aug 18, 2010 at 4:12 PM, Amanda Pinkston
<pinkston(a)fas.harvard.edu> wrote:
I receive the following error when running amelia on a
dataset with high
missingness:
Error in if (diff > iter.hist[count, 1] && count > 2) { :
missing value where TRUE/FALSE needed
This error has popped up several times, in different imputations (I'm trying
for five) and at different points of the EM chain. I can't get through five
imputations without this crash. What does the error mean?
Here is my code:
output <- amelia(data_1, cs = "country", ts = "year", intercs =
TRUE,
polytime = 2, p2s = 2, logs = c("gdpPPP2005", "gdppcppp",
"pop", "gnipc"),
ords = c("inc_group", "polxcons", "bnkv137",
"bnkv114", "iaeexbgt"), noms =
c("region", "legor"), empri = 0.01*nrow(data_1))
thanks,
Amanda Pinkston
-
Amelia mailing list served by Harvard-MIT Data Center
[Un]Subscribe/View Archive:
http://lists.gking.harvard.edu/?info=amelia
More info about Amelia:
http://gking.harvard.edu/amelia
-
Amelia mailing list served by Harvard-MIT Data Center
[Un]Subscribe/View Archive:
http://lists.gking.harvard.edu/?info=amelia
More info about Amelia:
http://gking.harvard.edu/amelia