Kim: If you are going to go the Monte Carlo resampling approach you will be better off going to compiled Fortran code than trying to implement it in S-Plus. For these kind of interative, looping proc
One possibility if there is a fairly good size sample (n > 100) with good coverage of the spatial extent is to fit a spatial trend surface by cubic polynomials of latitude and longitude coordinates i
Perhaps the important thing to keep in mind here is that our data can often deviate alot from normality yet inferences about means, changes in means, etc. will not be too far astray. But start asking
The regression quantile code for S-Plus written by Roger Koenker can be used to estimate and test for differences in any quantile for any contrast that might be incorporated in a linear model. Brian
I'm running S-Plus 2000 on a 1.4 ghz Pentium 4 with 512mb of RDRAM, Windows 2000 OS. I find that I cannot run S-Plus and a batch program (fortran 95 code for permutation tests) in the DOS window simu
Brian Ripley's interpretation seems correct to me. So the solution for a problem of this size (n = 100) is to make a lot more sample permutations (I suggest 10,000 for most analyses) to approximate t
This is related to a recent discussion about better ways to model data that are proportions. It has been suggested that rather than doing logit transformations on proportions and then treating them a
Dear S-Plus Users: I'm posting this again as I've had no response. The first part relates to model selection in glm when comparing different link functions for proportional data. The second part is a
A follow up on the estimate differences. I discovered that I was not changing tolerance and iterations correctly for S-Plus glm(). The default low value of epsilon=0.00001 and maxit=10 were not being
A follow up on the estimate differences. I discovered that I was not changing tolerance and iterations correctly for S-Plus glm(). The default low value of epsilon=0.00001 and maxit=10 were not being