I used the commands to increase R's RAM use (--max-mem-size=500 and --max-vsize=500M).  It still gives me the same error message, "Error: cannot allocate vector of size 109853 Kb" (except with a different number of Kb).  My other question is about how to change the time-series options in the windows version so that it does not use the cross-sectional variable & create dummy variables for all 8500 respondents).  This option is the global "_AMusecs=0;" in the online manual.  How do I implement this option in the windows version?

Kristin Burnett



On Tue, 03 Oct 2006 23:27:44 -0400 "Kristin Burnett" wrote:
-----Original Message-----
From: Levi Littvay (UNL) [mailto:https://webmail.psu.edu/webmail/main.cgi#] 
Sent: Tuesday, October 03, 2006 4:46 PM
To: KRISTIN D BURNETT
Subject: Re: [amelia] Fwd: amelia problem

Kristin

You need to increase the amount of RAM R uses.  I know of two ways of
doing this.  One is done when you run R.  Read about this at:

http://gking.harvard.edu/zelig/docs/How_do_I2.html

The other is by command:

memory.limit(1576)

(At least this is what I used.  I have 2GB of RAM, so you might want to
go lower.)

Try these and let me know how it worked.

Levi


Hello,

I have been trying to use the most recent Windows version of Amelia II for multiple imputation of my dataset (and I have the most recent windows version of R installed: 2.3.1).  My data have a max of 5 time points for about 8500 respondents (ends up with 33559 observation rows).  In the options, I identified the timepoint variable (ranges from 1 to 5) as the time series index and the person id # as the cross sectional index.   Then I identified which variables were nominal (which adds about 20 dummies altogether).  I left the TSCS options at the default.  And for priors I set boundaries of 0 to 1 on the dichotomous variables (about 5 of those), and set the boundary of 0 to 6 for one variable that I know should not extend beyond this range.  I am loading the data from Stata and requesting that the 5 output datasets also be in Stata format.

I keep getting an error message similar to the following:
<<<<<
There was an unexpected error in the execution of Amelia. 
Double check all inputs for errors and take note of the error message:

Error: cannot allocate vector of size 219444 Kb
>>>>>

I started by including all possible interactions, which involved a dataset of 92 variables.  But even after I took out the interaction variables and tried to run it with only 18 variables (although it is more like 38 if you account for the dummies created for nominal variables), and it still gave me the above error message.  This was also after increasing the workspace to 256MB (when the computer has 512MB of RAM).  Is the dataset just too large for the program?  Or am I doing something else wrong?  Thank you for whatever help you can give me.

Sincerely,
Kristin Burnett