I'm using IVEware to impute data for use of service facilities (includes 0-1 variables as well as many count variables). Since most of my variables are not normally distributed I'm trying to use the PERTURB statement which allows the choice of the Samping-Importance-Resampling algorithm for generating coefficients. This option never seems to take, though. I get the following error: Error: Repeated PERTURB statement ...despite having only one statement in the code. I considered that this might occur because of some implied PERTURB statement given the model I'm running, but I seem to get the error regardless (i.e., when I try to test it on a very simple model). Does anyone have any experience using this option in IVEware, or any idea whether a data issue might lead to this error? Any advice would be greatly appreciated. -Damon ========================================= Damon Jones, Ph.D., Research Associate Dept. of Health, Policy & Administration 116 Henderson Bldg. Pennsylvania State University University Park, PA 16802 Ph: 814-863-2908 Fx: 814-863-2905 -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.utsouthwestern.edu/pipermail/impute/attachments/20040507/d2bdeac1/attachment.htm From ugroempi <@t> ford.com Wed May 19 04:37:46 2004 From: ugroempi <@t> ford.com (Groemping, Ulrike (U.)) Date: Sun Jun 26 08:25:01 2005 Subject: [Impute] MIANALYZE: Variance in "Total covariance matrix" different from squared parameter standard error Message-ID: <[email protected]>
Hello everybody, I've just tried to get a total covariance matrix estimated by running MIANALYZE on the logistic regression estimates data set from five imputations. In the univariate output, one of my parameters of interest gets the estimated variance 0.031653, which corresponds to the standard error 0.177913. If I look at the Total covariance matrix (output when using the option MULT), the variance for the same parameter is 0.034983613, which corresponds to the standard error 0.187039. Does anyone have an explanation for this apparent mismatch? Any help would be greatly appreciated. Regards, Ulrike P.S.: Degrees of freedom for this parameter are 4761.8, fraction of missing information is 0.029391, just in case any of this information comes in useful (and by the way, I'm amazed how MIANALYZE knows this information, since it does not seem to be part of the ingoing SAS dataset of estimates and covariances; you probably note that I'm a beginner on MIANALYZE). Ulrike Groemping D-MC/4-B14 Ford-Werke AG D-50725 K?ln [email protected] +49-221/90-35666 Fax: -33021
