similar to: a question on statistics (rather than R-specific)

Dear R experts: I am running some calculations using polr() in MASS library, and found some differences in results obtained on two different machines (IRIX 6.5, and Linux RH 7.1). It is not clear to me whether this is due to some error in my programming the calculation and how to resolve the differences, if possible. The polr() call is the following:

It has always been my understanding that deviance for GLMs is defined by; D = -2(loglikelihood(model) - loglikelihood(saturated model)) and this can be calculated by (or at least usually is); D = -2(loglikelihood(model)) As is done so in the code for 'polr' by Brian Ripley (in the package 'MASS') where the -loglikehood is minimised using optim; res <-

With the incorporation of the useful hints, my user function now looks like this: userfn <- function() { ... ow <- options("warn") options(warn = 2); ... reg<-try(polr(act~.,data=mm,Hess=TRUE)) ... sumreg<-try(summary(reg)) print(length(sumreg)) print(sumreg) ... options(ow) # reset } The routine userfn() is called multiple times, two of which I happen to know to

Within an user function, how are the warnings from an R function be trapped (such that some proper actions can be taken)? 'last.warning' is returned only at the top level. Pointers are appreciated. -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.- r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html Send "info",

I'm running polr() and getting warning messages from glm.fit(). It seems reasonable to use glm.control() to turn on the trace and follow what glm.fit() does when called by polr(); or is it? glm.control(maxit=10, trace=TRUE) polr(act~., data=mm) The glm.control() sets the trace TRUE, but there's no change in the output from polr(). Many thanks in advance for any help/pointers.

Hi all, I am using polr(). The resulting coefficients of first levels are always 0. What to do if I wnat to get the coefficients of the last level 0. For example, suppose x has 3 levels, 1, 2, 3 probit <- plor(y ~ x, data1, method='probit') will get coefficients of level 2, 3 of x, but I want coefficients of level 1, 2 Thank you, Tian [[alternative HTML version deleted]]

Dear all, I've replicated the cheese tasting example on p175 of GLM's by McCullagh and Nelder. This is a 4 treatment (rows) by 9 ordinal response (cols) table. Here's my simple code: #### cheese library(MASS) options(contrasts = c("contr.treatment", "contr.poly")) y = c(0,0, 1, 7, 8,8,19, 8,1, 6,9,12,11, 7,6, 1, 0,0, 1,1, 6, 8,23,7,

I want to fit a multinomial model with logit link. For example let this matrix to be analyzed: male female aborted factor 10 12 1 1.2 14 14 4 1.3 15 12 3 1.4 (this is an example, not the true data which are far more complex...) I suppose the correct function to analyze these data is polr from MASS library. The data have been

I am running a windows executable, using wine, in a 'batch mode' - ie multiple times with different command parameters to the executable. For some parameters, the 'Program Error' dialog box appears and the wine process does not continue until the dialog box is closed. Is it possible to suppress the dialog box such that the entire batch can be completed without intervention?

Hi all, I'm working on a multinomial (or "polytomous") logistic regression using R and have made great progress using multinom() from the nnet library. My response variable has three categories, and there are two different possible predictors. I'd like to use the likelihoods of certain models (ie, saturated, fitteds, and null) to calculate Nagelkerke R-squared values for

Hi, I am searching for methods to compare regression models with an ordered categorical response variable (polr versus multinom). The pattern of predictions of both methods (using the same predictor variables) is quite different and the AIC is smaller for the multinom approach. I guess polr has more strict premises for the structure of the response variable, which methods can be used to test for

Dear all R-users, I am a new user of R and I am trying to build a discrete choice model (with more than two alternatives A, B, C and D) using logistic regression. I have data that describes the observed choice probabilities and some background information. An example below describes the data: Sex Age pr(A) pr(B) pr(C) pr(D) ... 1 11 0.5 0.5 0 0 1 40 1 0 0 0 0 34 0 0 0 1 0 64 0.1 0.5 0.2 0.2 ...

Hello, I am trying to run an ordinal logistic regression (polr) using the package 'MASS'. I have successfully run other regression classes (glm, multinom) without much problem, but with the 'polr' class I get the following error: " Error in svd(X) : infinite or missing values in 'x' " which appears when I run the "summary" command. The data file is

Hello, after calculating a multinomial logit regression on my data, I compared the output to an output retrieved with SPSS 18 (Mac). The coefficients appear to be the same, but the logLik (and therefore fit) values differ widely. Why? The regression in R: set.seed(1234) df <- data.frame( "y"=factor(sample(LETTERS[1:3], 143, repl=T, prob=c(4, 1, 10))), "a"=sample(1:5,

Hi I would like to analyse an ordinal categorical variable. I know how I can analyse a nominal categorical variable (with multinom or if there are only two levels with glm). Does somebody know which command I need in R to analyse an ordinal categorical variable? I want to describe the variable y with the variables x1,x2,x3 and x4. So my model looks like: y ~ x1+x2+x3+x4. y: ordinal factor

Dear Masters, first I'd like to wish u all a great 2011 and happy holydays by now, second (here it come the boring stuff) I have a question to which I hope u would answer: I run a logistic regression by glm(), on the following data type (y1=1,x1=x1); (y2=0,x2=x2);......(yn=0,xn=xn), where the response (y) is abinary outcome on 0,1 amd x is any explanatory variable (continuous or not)

Running a binary logit model on the data df <- data.frame(y=sample(letters[1:3], 100, repl=T), x=rnorm(100)) reveals some residual deviance: summary(glm(y ~ ., data=df, family=binomial("logit"))) However, running a multinomial model on that data (multinom, nnet) reveals a residual deviance: summary(multinom(y ~ ., data=df)) On page 203, the MASS book says that "here the

I want to get the logLik to calculate McFadden.R2 ,ML.R2 and Cragg.Uhler.R2, but the value from multinom does not have logLik.So my quetion is : is logLik meaningful to multinomial logistic model from multinom?If it does, how can I get it? Thank you! ps: I konw VGAM has function to get the multinomial logistic model with logLik, but I prefer use the function from "official" R

If one estimates a model using multinom, is it possible to perform the omnibus LR test ( the analogue to omnibus F in linear models ) using the output from multinom ? The residual deviance is there but I was hoping I could somehow pull out the deviance based on just using an intercept ? Sample code is below from the CAR book but I wasn't sure how to do it based on that example. Thanks

Dear R community, I am trying to write my own user defined split function for rpart. I read the example in the tests directory and I understand the general idea of the how to implement user defined splitting functions. However, I am having troubles with addressing the data frame used in calling rpart in my split functions. For example, in the evaluation function that is called once per node,