Hi Matt,

Hmm, that's odd.  The subset argument I used selected more than four-fifths of the very large data frame.  I checked the descriptives of the subset, and all of the variables vary.  Any ideas?  I appreciate your help.

Thanks.
-Isaac


On Thu, Nov 21, 2013 at 10:53 PM, Matt Blackwell <m.blackwell@rochester.edu> wrote:
Hi Isaac, 

It seems as though the subset you've chosen has some variables in it that do not vary. You might want to choose either a different or perhaps slightly larger subset to see if that works. 

Cheers,
Matt

~~~~~~~~~~~
Matthew Blackwell
Assistant Professor of Political Science
University of Rochester


On Thu, Nov 21, 2013 at 8:29 PM, Isaac Petersen <dadrivr@gmail.com> wrote:
Hi Matt,

I tried subsetting to keep a subset of rows.  Here's the error I received:

> overimpute(EXT_impute, var="extPOM", subset=tcid >= 12000)

error: inv(): matrix appears to be singular
error: chol(): failed to converge
terminate called after throwing an instance of 'std::runtime_error'
  what():

Any ideas?

Thanks!
-Isaac


On Thu, Nov 21, 2013 at 1:47 PM, Matt Blackwell <m.blackwell@rochester.edu> wrote:
Hi Isaac, 

One thing you can do is include a subset expression that refers to a column in the data. This might reduce the possibility of there being a mismatch between the dimensions of the subset vector and the original data. Here is an example: https://gist.github.com/mattblackwell/7587294[gist.github.com]

library(Amelia)
data(africa)
a.out <- amelia(africa, ts = "year", cs = "country")

overimpute(a.out, "gdp_pc")
overimpute(a.out, "gdp_pc", subset = country == "Burkina Faso")
overimpute(a.out, "gdp_pc", subset = gdp_pc >=500)

Hope that helps! If you're still seeing an error message, please send it along. 

Cheers,
Matt






On Thu, Nov 21, 2013 at 1:04 PM, Isaac Petersen <dadrivr@gmail.com> wrote:
I'm trying to run overimputation as a diagnostic on a large data set with 20 multiple imputations from Amelia.  The overimputation diagnostic on one variable takes a long time to complete (more than 2 weeks).  Is there any way to expedite the overimputation procedure?  For example, would using the subset argument to select a subset of cases expedite the procedure?  If so, how would one use the subset argument?  I tried using a vector of TRUEs/FALSEs to indicate whether to keep a row, but I received an error.  Any suggestions for expediting overimpute and/or using the subset argument of overimpute() would be very helpful.

Thanks!
-Isaac

--
Amelia mailing list served by HUIT
[Un]Subscribe/View Archive: https://urldefense.proofpoint.com/v1/url?u=http://lists.gking.harvard.edu/?info%3Damelia&k=p4Ly7qpEBiYPBVenR9G2iQ%3D%3D%0A&r=jLgdG6f%2BQq4pzHWI0S37ROhc5Jfy9q9oKEsPDdQXskc%3D%0A&m=ziKCXJVHystOQKDjsvoxN7IecJLp0WImP80a7ad0Cos%3D%0A&s=541b8f6aeff16535cc6e6f54635c10c0fa8d95cc9db513583b566fcb76b7650a
More info about Amelia: https://urldefense.proofpoint.com/v1/url?u=http://gking.harvard.edu/amelia&k=p4Ly7qpEBiYPBVenR9G2iQ%3D%3D%0A&r=jLgdG6f%2BQq4pzHWI0S37ROhc5Jfy9q9oKEsPDdQXskc%3D%0A&m=ziKCXJVHystOQKDjsvoxN7IecJLp0WImP80a7ad0Cos%3D%0A&s=be7ea59263860885c2717542ff8f04b09ededddc377730edf8b1dfa4812f3ff0
Amelia mailing list
Amelia@lists.gking.harvard.edu

To unsubscribe from this list or get other information:

https://urldefense.proofpoint.com/v1/url?u=https://lists.gking.harvard.edu/mailman/listinfo/amelia&k=p4Ly7qpEBiYPBVenR9G2iQ%3D%3D%0A&r=jLgdG6f%2BQq4pzHWI0S37ROhc5Jfy9q9oKEsPDdQXskc%3D%0A&m=ziKCXJVHystOQKDjsvoxN7IecJLp0WImP80a7ad0Cos%3D%0A&s=0993fc901107585586e8651c22c11938be61ddc1e5a73fbdda22901e86d9b250