similar to: Goodness-of-fit of GLM for Gamma Distribution

Hi All, could someone shed some light on what the difference between the estimated dispersion parameter that is supplied with the GLM function and the one that the 'gamma.dispersion( )' function in the MASS library gives? And is there consensus for which estimated value to use? It seems that the dispersion parameter that comes with the summary command for a GLM with a Gamma dist. is

I'm looking for goodness of fit tests for gamma distributions with large data sizes. I have a matrix with around 10,000 data values in it and i have fitted a gamma distribution over a histogram of the data. The problem is testing how well that distribution fits. Chi-squared seems to be used more for discrete distributions and kolmogorov-smirnov seems that large sample sizes make it had to

Thanks Petr. Comments below: At 03:40 PM 18/05/2007, Petr Klasterecky wrote: >Sean Connolly napsal(a): >>Hi all, >>I am wondering if anyone has written (or knows of) a function that >>will conduct a goodness-of-fit test for a gamma distribution. I am >>especially interested in test statistics have some asymptotic >>parametric distribution that is independent

Hi all, I am wondering if anyone has written (or knows of) a function that will conduct a goodness-of-fit test for a gamma distribution. I am especially interested in test statistics have some asymptotic parametric distribution that is independent of sample size or values of fitted parameters (e.g., a chi-squared distribution with some fixed df), because I want to fit gamma distributions to

I am making exploratory analyses on a complex survey data by using survey package. Could you help me how to see the goodness of fit for the model below? Should I use AIC, BIC, ROC, or what? What code would let me run a goodness of fit test for the model? Here are my codes: #incorporating design effects# > mydesign <- svydesign(id=~clust, strata=~strat, weights=~sweight, > data=mydata)

Hey, I make a regression for Gamma distribution with log link, in R and in SAS. In R, the dispersion is estimated by \phi=Deviance/(#_of_observations), In SAS, there are two options: \phi=Deviance/(#_of_observations-#_of_params) or \phi=Pearson/(#_of_observations-#_of_params). I understand that SAS formulae are correct, however, the coefficients, obtained by R and by the first version of SAS,

I am trying to test goodness of fit for my legalistic regression using several options as shown below.  Hosmer-Lemeshow test (whose function I borrowed from a previous post), Hosmer–le Cessie omnibus lack of fit test (also borrowed from a previous post), Pearson chi-square test, and deviance test.  All the tests, except the deviance tests, produced p-values well above 0.05.  Would anyone please

This is a summary and extension of the thread "GLMM (lme4) vs. glmmPQL output" http://maths.newcastle.edu.au/~rking/R/help/04/01/0180.html In the new revision (#Version: 0.4-7) of lme4 the standard errors are close to those of the 4 other methods. Thanks to Douglas Bates, Saikat DebRoy for the revision, and to G?ran Brostr?m who run a simulation. In response to my first posting, Prof.

Thank you for your reply. I am aware of the good reasons not to use the deviance estimate in binomial, Poisson, and gamma families. However, for the inverse Gaussian, the choice seems to me less clear cut. So I just wanted to compare two different options. I have used the dispersion parameter to compute the standardized deviance residuals: summary(model.gamma)$deviance.resid

> Hello, > > I have a question about a discrepancy between the > reported F statistics using anova() and add1() from > adding an additional term to form nested models. > > I found and old posting related to anova() and > drop1() regarding a glm with a dispersion parameter. > > The posting is very old (May 2000, R 1.1.0). > The old posting is located here. >

Dear R users, Which test is most appropriate in glm when the family is Gamma? In the help page of anova.glm, I found the following ?For models with known dispersion (e.g., binomial and Poisson fits) the chi-squared test is most appropriate, and for those with dispersion estimated by moments (e.g., gaussian, quasibinomial and quasipoisson fits) the F test is most appropriate.? My questions :

Apologies, basic question on plot. y <- c(-4,3,-2,1); x <- c("time 1", "time 2", "time 3", "time 4"); plot(x,y, type="b"); of course fails. x <- 1:4 makes it succeed, but then I have too many ticks on my X axis. I want exactly 4 tickmarks. It would also be nicer if I could name the ticks. I looked at ?par and Venables&Ripley,

Dear R users, Overdispersion is often a problem in binomial data. I attempt to model a binary response (sex-ratio) with three categorical explanatory variables, using GLM, which could assume the form: y<-cbind(sexf, sample-sexf) model<-glm(y ~ age+month+year, binomial) summary(model) Output: (Dispersion parameter for binomial family taken to be 1) Null deviance: 8956.7 on 582

Hello all, I was wondering if someone can enlighten me as to the difference between the logLik in R vis-a-vis Stata for a GLM model with the gamma family. Stata calculates the loglikelihood of the model as (in R notation) some equivalent function of -1/scale * sum(Y/mu+log(mu)+(scale-1)*log(Y)+log(scale)+scale*lgamma(1/scale)) where scale (or dispersion) = 1, Y = the response variable, and mu

Hi Bianca, I hope you?ve solved your problem with SPSS and R probit analysis, but if you haven?t, I have your solution: Based on the output you?ve given, I see that your residual deviance is under-dispersed (that the ratio of residual deviance to residual deviance df does is less than 1). However, you?ve told R to treat your dispersion parameter as 1 (you did this by using the ?family =

Hi, I would like to obtain the dispersion parameter for a quasipoisson model for later use in calculating QAIC values for model comparison.Can anyone suggest a method of how to go about doing this? The idea I have now is that I could use the residual deviance divided by the residual degrees of freedom to obtain the dispersion parameter. The residual deviance is available in the summary

THANKS so very much for your help (previous and future!). I have a two follow-up questions. 1) You say that dispersion = 1 by definition ....dispersion changes from 1 to 13.5 when I go from binomial to quasibinomial....does this suggest that I should use the binomial? i.e., is the dispersion factor more important that the 2) Is there a cutoff for too much overdispersion - mine seems to be

Dear all, I've got some questions, probably due to misunderstandings on my behalf, related to fitting overdispersed binomial data using glm(). 1. I can't seem to get the correct p-values from anova.glm() for the F-tests when supplying the dispersion argument and having fitted the model using family=quasibinomial. Actually the p-values for the F-tests seems identical to the p-values for

I'm confused as to the trustworthiness of the dispersion parameters reported by glm. Any help or advice would be greatly appreciated. Context: I'm interested in using a fitted GLM to make some predictions. Along with the predicted values, I'd also like to have estimates of variance for each of those predictions. For a Gamma-family model, I believe this can be done as Var[y] =

I am running a binomial glm with response variable the no of mites of two species y->cbind(mitea,miteb) against two continuous variables (temperature and predatory mites) - see below. My model shows overdispersion as the residual deviance is 48.81 on 5 degrees of freedom. If I use quasibinomial to account for overdispersion the dispersion parameter estimate is 2501139, which seems