If you get an r.dll error I assume you are running the Windows version of R. The Hmisc library for Windows needs to be updated to incorporate the latest version of aregImpute. This will be done in a couple of weeks. The latest version of Hmisc for R is available for Linux/Unix/MacOSX at http://hesweb1.med.virginia.edu/biostat/s/Hmisc.html . This version has the up-to-date online documentation, in which you will see that I added another variable, x3, to the examples. That is because predictive mean matching does not work with fewer than 3 variables used to predict the target variable. In the extreme case of one right-hand-side variable and assuming that only monotonic transformations of left and right-side variables are allowed, every bootstrap resample will give predicted values of the target variable that are monotonically related to predicted values from every other bootstrap resample. The same is true for Bayesian predicted values. This causes predictive mean matching to always match on the same donor observation. It took me a lot of pain to figure this out. This is now discussed in the documentation.
Frank On Wed, 17 Apr 2002 01:35:26 +1200 nmi13 <[EMAIL PROTECTED]> wrote: > Dear Dr. Frank, > > I tried to use the aregImpute, with the example given in your manual and when > the whole program is run it given an error r.dll. Can you please suggest me > why this is occurring. when I tried with a less number of iterations and the > two variables there is no error. It gives the results. One more doubt with the > program. > > I created a data as follows > x1<-1:100 > x2<-x1+rnorm(100) > orgi.x2<-x2[1:30] > x2[1:30]<-NA > and now used the aregImpute to impute the missing values > f<-(~I(x1)+I(x2),n.impute=5) > now when I check the imputed value it is just the 31st value of teh data which > is obtained as imputed value, 8 times and replaced in for the first 30 values > missing. When the original values are compared to the imputed there is huge > difference to them and the imputed ones. I thought I might be wrong and I did > a very small example this time > with the following program > x<-1:10 > y<-x+rnorm(10) > orgi.y<-y[1:3] > y[1:3]<-NA > f<-aregImpute(~y+x,n.impute=5) > now checked the values to the original ones and found the same results as > above. Am I doing some thing wrong or the package is meant to give the results > the same way I don't know. Can You please correct me if I am wrong and correct > if I am doing the wrong procedures? If I am correct then can you please > explain the reason for only coming up with the first observed value as the > imputed value. Thank you very much for your time and help.Even I tried with > the small example given in your package thinking that I might be worng in > creating the data set in a wrong way, but for this too the vales are the same > as the first observed value in the data set. > Sincerely yours > Murthy. > > M.N.Murthy > Student > Department of Mathematics and Statistics > University of Canterbury, > New Zealand. > ph: 0064-3-3411500 extn 52193 > > -- Frank E Harrell Jr Prof. of Biostatistics & Statistics Div. of Biostatistics & Epidem. Dept. of Health Evaluation Sciences U. Virginia School of Medicine http://hesweb1.med.virginia.edu/biostat
