similar to: standard errors from glm (PR#3180)

G'day all, I had a look at the GLM code of R (1.4.1) and I believe that there are problems with the function "glm.fit" that may bite in rare circumstances. Note, I have no data set with which I ran into trouble. This report is solely based on having a look at the code. Below I append a listing of the glm.fit function as produced by my system. I have added line numbers so that I

Here's a patch that should make predict.glm(..., type="response") retain the names. The change passes make check on our Opteron running SLES9. One simple test is: names(predict(glm(y ~ x, family=binomial, data=data.frame(y=c(1, 0, 1, 0), x=c(1, 1, 0, 0))), newdata=data.frame(x=c(0, 0.5, 1)), type="response")) which gives [1]

Dear R-Users, I wish to create a new family object based on the Binomial family. The only difference will be with the link function. Thus instead if using the 'logit(u)' link function, i plan to use '-log(i-u)'. So far, i have tried to write the function following that of the Binomial and Negative Binomial families. The major problem i have here is with the definition of the

I think one needs an LU decomposition rather than QR. However, I couldn't find anything off the shelf to do an LU, other than learning that determinant() now uses LU instead of QR or SVD, so the code to do it must be in there for those that want it. You'll probably need to divide rows of U by the first entry if you insist on the unique reduced REF. However, I can't see any reason

Greetings, In my opinion the documentation or behaviour of qr.solve, qr.coef, qr.resid, and qr.fitted is not easily comprehensible and unfortunate. We all know that a linear system Ax=b can have 0, one or infinitely many solutions. To treat all these cases uniformly we can rephrase the problem as x = argmin_u||Au-b||,

Below is a SIMEX object that was generated with the "simex" function from the "simex" package applied to a logistic regression fit. From this mountain of information I would like to extract all of the values summarized in this line: .. ..$ variance.jackknife: num [1:5, 1:4] 1.684 1.144 0.85 0.624 0.519 ... Can someone suggest how to go about doing this? I can extract the

Full_Name: Christian Brechbuehler Version: 2.4.1 Patched (2007-03-25 r40917) OS: Linux 2.6.15-27-adm64-xeon; Ubuntu 6.06.1 LTS Submission from: (NULL) (24.61.47.236) Splus and R have different ideas about what qr.coef(qr()) should return, which is fine... but I believe that R has a bug in that it is not internally consistent, and another separate bug in the documentation. In particular, on

On Thu, 19 Apr 2007, brech at delphioutpost.com wrote: > Full_Name: Christian Brechbuehler > Version: 2.4.1 Patched (2007-03-25 r40917) > OS: Linux 2.6.15-27-adm64-xeon; Ubuntu 6.06.1 LTS > Submission from: (NULL) (24.61.47.236) > > > Splus and R have different ideas about what qr.coef(qr()) should return, > which is fine... but I believe that R has a bug in that it is not

Hi, I wish to write a new link function for a GLM. R's glm routine does not supply the "loglog" link. I modified the make.link function adding the code: }, loglog = { linkfun <- function(mu) -log(-log(mu)) linkinv <- function(eta) exp(-exp(-eta)) mu.eta <- function(eta) exp(-exp(-eta)-eta) valideta <- function(eta) all(eta != 0)

I know this has been dealt with before on this list, but the previous messages lacked detail, and I haven't figured it out yet. The model is: \x_{ij} = \mu + \alpha_i + \beta_j \alpha is a random effect (subjects), and \beta is a fixed effect (condition). I have a link function: p_{ij} = .5 + .5( 1 / (1 + exp{ -x_{ij} } ) ) Which is simply a logistic transformed to be between .5 and 1.

Hi all, I wonder if this kind of questions are ok in this list... Quick question: What does it mean than the rank of the QR decomposition of a NxN matrix is N-1 ? m: NxN matrix qr(m)$rank equal to (N-1) Long version: I'm doing a manova on a matrix of 10 variables and 16 observations. > dim(tmp) [1] 16 10 > fit <- manova( tmp ~ treatment*mouse ) >results <-

I suggest adding a 'pivot' argument to qr.R, to obtain columns in the same order as the original x, so that a <- qr(x) qr.Q(a) %*% qr.R(a, pivot=TRUE) returns x. -------------------------------------------------- # File src/library/base/R/qr.R qr.R <- function(qr, complete = FALSE, pivot = FALSE) { # Args: # qr: a QR decomposition, produced by qr() # complete:

FWIW p. 206 of the White Book gives the following for names(binomial()): family, names, link, inverse, deriv, initialize, variance, deviance, weight. So $aic wasn't there In The Beginning. I haven't done any more archaeology to try to figure out when/by whom it was first introduced ... Section 6.3.3, on extending families, doesn't give any other relevant info. A patch for

I need to perform a QR-decomposition of a sparse matrix, so I've been trying to use the Matrix package. Unfortunately I don't seem to be getting exactly the same results as if I had used the qr() command from the base package. Here is an example of what I'm doing. > spdata <-rpois(50,1) > y <- rnorm(10,0,1) > S <-

I was noticing mainly sign differences amoung the solutions to QR decomposition. For example R: > x <- matrix(c(12,-51,4,6,167,-68,-4,24,-41),nrow=3,byrow=T) > x [,1] [,2] [,3] [1,] 12 -51 4 [2,] 6 167 -68 [3,] -4 24 -41 > r <- qr(x) > r$qr [,1] [,2] [,3] [1,] -14.0000000 -21.0000000 14 [2,] 0.4285714 -175.0000000 70 [3,]

Dear All, Is there is a fast way of refitting lm() when the design matrix stays constant but the response is different? For example, y1 ~ X y2 ~ X y3 ~ X ...etc. where y1 is the 1st instance of the response vector. Calling lm() every time seems rather wasteful since the QR-decomposition of X needs to be calculated only once. It would be nice if qr() was called only once and then the same

Hi all, Regular and avid readers of this column will know that Don Driscoll and I have recently posted two messages requesting assistance concerning an apparent failure of "se.contrast" to produce an se for a contrast. So far, an ominous silence rings in our ears, but read on Gentle Reader, and see if even the machinations of "debug" doesn't stimulate you to respond with a

Here is a modified version of qraux.Rd, an edited version of R-2.14.0/src/library/base/man/qraux.Rd This gives some details and an example for the case of pivoting. In this case, it is not true that X = QR; rather X[, pivot] = QR. It may save some other people bugs and time to have this information. Tim Hesterberg -------------------------------------------------- % File

At useR 2006 I mentioned that it would be nice to have a way to specify binomial links that involved free parameters and described some experience with a Gosset link involving a free degrees of freedom parameter, and a Tukey-lambda link with two free parameters. My implementation of this involved some rather kludgey modifications of binomial, make.link and glm that (essentially) added a

John, You have specified a model with E(y) = 1/eta where eta = X beta is the linear predictor and E(y) must be >0, since the family is Gamma and you have a lot of covariates in the model. glm now has to try to find a best linear predictor, but under the constraint that eta>0 for every single one of the observations (the log-likelihood involves a log(eta) term). The internal