 |
 |
I'm attempting to use Bugs to do multiple imputation in the context of multivariate regression, and I've come up with a trap zero error. Though there are a couple mentions of this in the list archives, none of them seem appropriate to my situation. The idea is that I have several variables which are fully observed, and a number which are partially observed. I estimate the partially observed variables one at a time using the fully observed variables, and then move this estimate to the right side of the equation to help estimate the next partially observed variable. Ultimately, I use all of the variables to estimate the quantity I'm really interested in--in this case, an index of social capital. I'm relatively new to BUGS, so it's quite possible I'm missing something. But the model compiles fine, loads the data set, and accepts the initial values for the regression coefficients (I computed these by a standard regression using listwise deletion). The error message appears when I try to generate the remaining initial values--the initial values for the data points I'm missing. I've included my code, as well as the error message. I'm on my own with this, as I'm the only one in my department (politics) who does anything with BUGS. Any help you can offer is *greatly* appreciated. With thanks, Matt