similar to: 1.3.1/1.4.1 Windows binary incompatibilities?

I'm probably missing something very basic here, but: I've written some C code that I load into R dynamically. In the course of this C code, I generate some multinomial random deviates. I initially used the publically available "randlib" library, which also implements its own random number generator and binomial deviates (which are used to generate the multinomial deviates).

OK, a boneheaded question ... I've made a set of packages for my students. I'm serving these packages from my web site in the form of a set of tar.gz source packages (constructed with R CMD build) and a set of .zip Windows binary packages, constructed by cross-compiling according to Brian Ripley's instructions (make pkg-foo in the src/gnuwin32 directory) and then zipping up the

I have a general style question about R coding. Suppose I'm writing a function (foo1) that calls other functions (foo2, foo3, ...) which have complicated argument lists (e.g. optim(), plot()), _and_ I may be calling several different functions in the body of foo1. Since foo2 and foo3 have different sets of arguments, I can't just use "..." ; I did write some code a while ago

I'm very excited by the new mle package now incorporated in stats4. If possible, I'd like to help develop it. In the past I wrote a similar package (mleprof, available from http://www.zoo.ufl.edu/bolker/R/src), and would like to see if there's anything that my package does that I could contribute (in particular, I'd like to make sure that the code is as robust as possible in

I have made a minor hack to "legend" (in R 1.0.0, but I didn't notice any changes to legend in the 1.0.1 NEWS) to allow the legend to be formatted in multiple columns, or horizontally (number of columns <- number of legend items). (I find this helpful when I have lots of legend items and not a lot of vertical space to squeeze the legend into.) (Another hack I've considered

I probably didn't follow the discussion of allowing "=" as an assignment operator closely enough, but I was a little bit horrified to discover today (using 1.6.0; I haven't upgraded to 1.6.1 yet) that x <- runif(20) y <- 1:20 y[x=min(x)] gives numeric(0) (because min(x) is non-integer). x <- sample(1:20,20,TRUE) y[x=min(x)] is even worse -- it gives the

PR #7690 points out that if the confidence intervals (+/-1.58 IQR/sqrt(n)) in a boxplot with notch=TRUE are larger than the hinges -- which is most likely to happen for small n and asymmetric distributions -- the resulting plot is ugly, e.g.: set.seed(1001) npts <- 5 X <- rnorm(2*npts,rep(3:4,each=npts),sd=1) f <- factor(rep(1:2,each=npts)) boxplot(X~f) boxplot(X~f,notch=TRUE) I can

I'm trying to answer a question from a student about memory use in R (I won't go into the details right here). I have a really vague memory of having read a document, possibly by Venables or Ripley, discussing the awfulness of memory allocation in S-PLUS, and giving (in the context of a bootstrapping analysis of shoe size data??) some general strategies for conserving memory in S-PLUS.

Not really a bug (I eventually discovered), but ... When running GaussRF()/DoSimulateRF() with n>1, the default pch="!" prints backspaces to the terminal, which causes problems (LaTeX barfs) if you're running this within an Sweave() job. Setting RFparameters(pch="") fixes the problem. Could put this in the documentation, but I don't know where you could

Any better (more efficient, built-in) ideas for computing coef[1]+coef[2]*x+coef[3]*x^2+ ... than polynom <- function(coef,x) { n <- length(coef) sum(coef*apply(matrix(c(rep(x,n),seq(0,n-1)),ncol=2),1,function(z)z[1]^z[2])) } ? Ben -- 318 Carr Hall bolker at zoo.ufl.edu Zoology Department, University of Florida http://www.zoo.ufl.edu/bolker

When using test="F" in stat.anova() / anova.glm(), R uses the assumed dispersion parameter for the specified family (e.g. scale=1 for binomial), while S-PLUS automatically uses the estimated dispersion parameter (residual deviance/residual df). I think there are good reasons for the behavior in R -- it fits with the "you get what you actually asked for" philosophy -- and

For various reasons, I've opted to make my packages available from my own web page rather than submitting them to CRAN (mostly laziness -- for a long time I didn't have the packages quite cleaned up enough to pass all the tests). It occurred to me to wonder about the scripts used by CRAN maintainers to generate the PACKAGES file, and to generate PACKAGES.html from PACKAGES. Are

I was sitting in the coffee room at work listening to people complain about a recent seminar about nanotechnology using the terms nanofluidics, nanofactory, nano-this, and nano-that ... I found myself thinking "well the speaker should just have said with(nano, ...) Un(?)fortunately there's no-one here I can share that thought with. -- 620B Bartram Hall

I appreciate Brian and Martin's answers -- and I certainly don't spend as much time & energy maintaining and answering questions about R as they do -- *but* it does seem to me that it would make a number of new (switching) user's lives easier if there were a succinct list of these differences, with a disclaimer ... I would be willing to maintain such a list, but since I

I'm trying to do a simple (?) analysis of a 1D spatial data set, allowing for spatial autocorrelation. (Actually, I'm comparing expected vs. observed for a spatial model of a 1D spatial data set.) I'm using models like gls(obs~exp,correlation=corExp(form=~pos),data=data) or gls(obs~exp,correlation=corLin(form=~pos),data=data) This form is supposed to fit a linear model of

In the optim() documentation, the control parameter "maxit" says "There is no other stopping criteria." That should be "are", or "criterion" ... While I'm at it -- poking at the code a little bit, it looks as if the initial simplex is set from the initial point by displacing each parameter value by max(0.1,0.1*pmax(fabs(Bvec))), where Bvec is the

I'm trying to figure out the best way of fitting the same negative log-likelihood function to more than one set of data, using mle() from the stats4 package. Here's what I would have thought would work: -------------- library(stats4) ## simulate values r = rnorm(1000,mean=2) ## very basic neg. log likelihood function mll <- function(mu,logsigma) {

The besselK() function knows about overflows/underflows internally; there is a constant xmax_BESS_K in src/nmath/bessel.h (and referred to only in bessel_k.c), equal to 705.342, which is checked if expon.scaled is FALSE. (The equivalent number for bessel_i.c is 709, defined as exparg_BESS in bessel.h.) However, besselK(x) silently returns +Inf if x>705.342. This behavior is reasonable for

I just sent a bug report about "mailx" being the default mailer, before I looked at bug.report() carefully and saw that it checked "/usr/ucb/mail", and "Mail" as well as "mailx". I guess I had had the problem with this in a previous version. Would it be clearer to have .Options$mailer be a vector or list of alternate mailers that were checked in turn,

Just in case anyone cares or is hitting the same problem: to install current mgcv (1.0-5) on 1.9.0 on Knoppix/Debian unstable I had to: # cd /usr/lib # ln -s /usr/lib/atlas/libblas.so.3 libblas-3.so # ln -s /usr/lib/atlas/liblapack.so.3 liblapack-3.so Otherwise compilation couldn't find -lblas-3 or -llapack-3 (I could have gotten away with the links in /usr/lib/atlas instead of /usr/lib,