Hi Jianhe, you probably will need to do some reading on tree models to be able to use them correctly and efficiently. I suggest the documentation to the library rpart, which you can find by googling
Dear list members, I created 110 slm() models (family CAR) and wanted to test them on reserved data. However, it seems that no method is implemented in slm() for predicting onto new data. I can under
Dear list members, I want to create a 3D persp plot that does not cover a square. Is that possible? I calculated predictions from a model for certain parameter combinations of R (growth rate) and dea
Thank you to Brian Ripley and Ray Brownrigg. Using NAs to fill the rectangle worked. I misunderstood the help file. However, this solution leads to "jagged" edges where the graph is cut off by NAs (s
Dear list members, I received the following error message while running a likelihood ratio test on spatial slm() models (with module(spatial)). I have used this technique before successfully, so I'm
I found my error: despite my assurance that I checked everything, I passed a character vector instead of a numerical vector to the coefficients argument. I would have never found that through the err
Dear list members, I run and prune rpart trees in a batch file and occasionally receive a strange error that leads to a failure in the batch execution. However, the failure does not happen predictabl
Thanks. I did read the help page but simply did not think that there could be more than 1 minimum given the many digits Splus calculates. And because cross-validation is random I didn't get the effec
Dear all, I'm trying to install the rpart library from Brian Ripley's webpage (http://www.stats.ox.ac.uk/pub/S/S+6/) on my Linux 7.2 computer (system information below). Although I followed the instr
Dear list members, has anyone implemented a function that automates pruning rpart regression trees according to the 1 SE rule? I searched the s-news archives but couldn't find anything posted along t
Dear all, I implemented my own request for an automated 1 SE pruning in rpart with the following command: + min(test$cptable[,4])+test$cptable[test$cptable[,4]==min(test$cptable[,4]),5 ],1])) where "
I interpreted Burnham and Anderson's formula as: Loglik = (-n / 2) * LN(RSE^2 * df / n) - (n / 2) * LN(2*PI()) - (n / 2) With RSE = residual standard error (in output) df = degrees of freedom on the
You can export your data as a comma-speparated file or a dbase format file and import it into Arcview. Note that the variable name conventions are different in ArcView, most noticably it won't accept
I believe the paper below did a comparison relevant to your question but possibly not exactly what you are asking (I'm not sure that nnet is artificial neural networks - on of the techniques compared