RE: [amelia] CSTS data with boundary priors

Hi Matt and Wayne, Thanks for your help. I updated to the very latest version (1.2-12). But I am still having trouble with the bounds priors. This is the message I am getting: Error in AMr1[, varPriors, drop = FALSE] : incorrect number of dimensions Nonetheless, I will follow Matt's advice and work with transformed variables instead. Without the bounds, everything seems to be working fine. Best, Milan -----Original Message----- From: Matt Blackwell [mailto:mblackwell@gmail.com] Sent: Wednesday, July 15, 2009 9:42 AM To: Milan Svolik Cc: amelia(a)lists.gking.harvard.edu Subject: Re: [amelia] CSTS data with boundary priors Hi Milan, Are you receiving an error message from Amelia? Or are the imputation falling outside of the specified bounds? In addition to the bounds argument, there is another way to specify two-sided bounds using transformations. If you can transform your variable into a (0,1) variable (that is, a proportion between the two bounds), then you can pass this transformed variable to the logistic argument (lgstc), which will force Amelia to always impute it between 0 and 1. After you get the output back from Amelia, you can put it back on to the original scale with code like this: for (i in 1:a.out$m) { a.out$imputations[[i]]$myvar <- scale*a.out$imputations[[i]]$myvar + offset } Filling in the scale and offset values according to the transformation to a (0,1) variable. This might be of help to someone trying to implement bounds. In any case, please give us a bit more detail on your problem and we can try to run it down. Thanks, matt. On Wed, Jul 15, 2009 at 9:13 AM, Milan Svolik&lt;msvolik(a)gmail.com&gt; wrote: update problems - 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