similar to: Question about GAM (mgcv)

The plots look reasonable to me. The plot of residuals against linear predictor always looks scary when many of the fitted values are very close to zero, so I tend to look at residuals against sqrt(fitted) in such cases. I don't think that the presence of the zero curve is a reason to reject the model --- it's easy to produce such plots by fitting a completely correct model to simulated

>> I have come across your reported log1p error (#2837) on a NetBSD (1.6W) >> system. I've just made further experiments on the deficient log1p() function on OpenBSD 3.2 and NetBSD 1.6 with this test program: % cat bug-log1p.c #include <stdio.h> #include <stdlib.h> #include <math.h> int main(int argc, char* argv[]) { int k; double x; for (k = 0; k

Brian Ripley writes today: >> There is already a usable log1p implementation in src/nmath/log1p, for >> platforms without it. All we need to do is to arrange to use it on those >> systems with broken versions. That's not easy without access to such a >> platform to test it, though. I need the same kind of test in my own software, so I made some experiments and

Ray Brownrigg <ray@mcs.vuw.ac.nz> writes today about the inaccurate log1p() on NetBSD 1.6 and OpenBSD 3.2: >> Well, the source which I have access to doesn't bear that out. Interesting. Our NetBSD installation is pretty recent: % uname -a NetBSD netbsd.vm.math.utah.edu 1.6 NetBSD 1.6 (GENERIC) #0: Sun Sep 8 19:43:40 UTC 2002

I have a 30 x 27 data frame, which I'm trying to scale and transform. Only I want to scale certain all variables except one (the dependent, which I want to log+1 transform). I can use split() and then scale() and log1p(), but I'm wondering if I can do this in one call. I tried apply(), but I could only get the whole data frame, not a part of it: > test <- apply(chin.sub, 2, log1p)

G'day all, I am working on problems where I have to calculate the logarithm of a sum or difference from the logarithms of the individual terms; so the functions logspace_add and logspace_sub which are part of R's API come in handy. However, I noticed that logspace_sub can have problems if both arguments are (very) small or the difference between the arguments are vary small. The logic

Hi, pbirthday(, coincident = 2) starts to issue warnings (see (*) below) for larger number of classes (R 4.0.0, R-devel ./src/library/stats/R/birthday.R:47). The default coincident = 2 is computed as 1 - prod((c:(c - n + 1))/rep(c, n)) where c = classes. Using exp(log(...)), one can derive the return value if(n > 0) 1 - exp(sum(log1p(-(0:(n-1))/c))) else 0. Simplifying this a bit further one

Ray Brownrigg <ray@mcs.vuw.ac.nz> writes today about the log1p() problems on NetBSD 1.6 and OpenBSD 3.2: >> Indeed, but since I have acess to a NetBSD developer, this is an >> opportunity to get it fixed for everyone. That's a start, but the reality is that it isn't enough. There will be many sites that don't upgrade O/S versions except at possibly very long

I've looked around and I just can't find anything that will work for my needs. This is a bit of a 2 part question but pertaining to the same topic so bare with me. The first is with my qq plot. On the Y axis of my qq plot it'll have my sample quantities but because my data is log-normal it'll show numbers between 0 - 5 (depending on the data). I'd like to know how to get it,

There is a small typo in the Source section of the help page for log1p: Source: 'log1p' and 'expm1' may be taken from the operating system, but if not available there are based on the Fortran subroutine 'dlnrel' there -> they Jen > sessionInfo() R version 3.0.2 (2013-09-25) Platform: x86_64-unknown-linux-gnu (64-bit) locale: [1] LC_CTYPE=en_US.UTF-8

Dear People, I am trying to use R's Math library as standalone, as documented in /src/nmath/standalone. I am using C++ in Debian testing, and the versions are as follows: ii g++-3.0 3.0.3-1 The GNU C++ compiler. ii r-mathlib 1.4.0-1 `GNU S' - Standalone R math library I have a file (rand.cc) as follows. I don't think that lattice.hh or mh.hh are very

I've now done two rebuilds of R-1.7.0 on NetBSD 1.6, one with the --without-zlib configure option, and one without. Both builds use the recently-installed gcc-3.2.3 compiler. As before, the one built normally gets a segment violation, whereas the one built with the --without-zlib option works. The odd thing is that neither uses shared libraries for zlib: % ldd /usr/local/lib/R/bin/R.bin

I have hit a problem with the design of the mcmc package I can't figure out, possibly because I don't really understand the R function call mechanism. The function metrop in the mcmc package has a ... argument that it passes to one or two user-supplied functions, which are other arguments to metrop. When the two functions don't have the same arguments, this doesn't work.

The issue is that without an extensible derivative table or the proposed extensions, it is not possible to automatically produce (without manual modification of the deriv3 output) a function that avoids catastrophic cancellation regardless of the working range. Manual modification is not onerous as a one-time exercise, but can be time consuming when it must be done numerous times, for example

Dear list, i'm checking the residuals plots of a gam model after a processus of model selection. I found the "best" model, all my terms are significant, the r-square and the deviance explained are good, but I have strange residuals plots: http://dl.dropbox.com/u/1169100/gam.check.png http://dl.dropbox.com/u/1169100/residuals_vs_fitted.png What does explains the "curve"

Hi list, I tried to install MCMCpack to R-2.12.0, got the following error, CC -m64 -library=stlport4 -I/apps/sparcv9/R-2.12.0/lib/R/include -DSCYTHE_COMPILE_DIRECT -DSCYTHE_DEBUG=0 -DHAVE_TRUNC -DHAVE_IEEEFP_H -I/opt/csw/include -KPIC -g -c MCMCSVDreg.cc -o MCMCSVDreg.o "error.h", line 598: Error: The function "abort" must have a prototype.

Hallo I need some help with the output provided by gam.check after a gam fit (using the package mgcv). To give a brief description of my data, I have claims: a vector of values, which include NA's and one large negative value - otherwise all positive (55 values in total that are not NA). origin: a factor with 10 levels j : taking the values 1, 2, ...., 10 I have fitted a gam, with: >

R-users E-mail: r-help@r-project.org Hi! R-users. A simple object as below was created to see how gam() of package "mgcv" and anova() work. function() { library(mgcv) set.seed(12) nd <- 100 xx1 <- runif(nd, min=1, max=10) xx1 <- sort(xx1) yy <- sin(xx1)+rnorm(nd, mean=5, sd=5) data1 <- data.frame(x1=xx1, y=yy) fit1 <- gam(y~s(x1, k=5),

About April 25th, there had been some changes within R-devel's src/nmath/pnbeta.c (and probably some other relevant places) and now building R-devel on FreeBSD 10.0-CURRENT (amd64) with gcc-4.6.4 and math/R-devel (selfmade forked port from math/R) fails like this: [..snip..] mkdir /usr/ports/math/R-devel/work/R-devel/bin/exec gcc46 -std=gnu99 -I../../src/extra -I. -I../../src/include

Hi. Unless I'm misremembering, log, exp, sin, cos, and tan are all handled in deriv3. The functions listed are specially coded slightly more accurate versions but can be substituted with native ones for which deriv/deriv3 will work automatically. I believe that if you write your functions using log(a + 1) instead of log1p(a) or log(x) / log(2) instead of log2(x) deriv3 will work fine.