similar to: [RfC] Family dispersion

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

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

Following p.206 of "Statistical Models in S", I wish to change the code for summary.glm() so that it estimates the dispersion for binomial & poisson models when the parameter dispersion is set to zero. The following changes [insertion of ||dispersion==0 at one point; and !is.null(dispersion) at another] will do the trick: "summary.glm" <- function(object, dispersion =

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

>>>>> Ben Bolker >>>>> on Sun, 3 Jun 2018 17:33:18 -0400 writes: > Is it generally known/has it been previously discussed here that the > $aic() component in GLM-family objects (e.g. results of binomial(), > poisson(), etc.) does not as implemented actually return the AIC, but > rather -2*log-likelihood + 2*(model_has_scale_parameter)

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] =

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

Dear list, This is a question about overdispersion and the ML estimates of the parameters returned by the glm.poisson.disp (L. Scrucca) and glm.nb (Venables and Ripley) functions. Both appear to assume a negative binomial distribution for the response variable. Paul and Banerjee (1998) developed C(alpha) tests for "interaction and main effects, in an unbalanced two-way layout of counts

I am trying to know exactly the formulas used to calculate aic for glm models. In glm.fit, the calculus of aic is: aic.model <- aic(y, n,mu, weights, dev) + 2 * fit$rank where 2 * fit$rank is (may be am i wrong?) twice the numbers of parameters p and aic(y, n, mu, weights, dev) refers to the function defined in the family function (which is for Gamma family, for instance) aic

Dear all, I am having difficulty interpreting the parameter estimates from the stablereg function. Specifically I am trying to keep things simple to start with by using stablereg to fit a normal distribution to a simulated data set from that distribution (in order to understand the way that stablereg reports parameter estimates). I cannot work out the scale on which the dispersion parameter (some

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 :

Hello List, I'm building a package for R, and I want to use the font family 'symbol', because it seems to be the most consistent of the four font families that par guarantees. The problem is that when I tried to run it on a different machine, I got an X11 error message saying that that font family was not available. What I would like to do, therefore, is do a check at the beginning

--Multipart_Wed_Feb__4_12:25:40_1998-1 Content-Type: text/plain; charset=US-ASCII Jim, I am relating your message to R-devel. This should be discussed in a broader audience; I am not an expert on GLM's, I know you are and others on this group also... R-develers, please CC to Jim Lindsey (on this topic), since he hasn't been part of the R-devel list for a while.. BTW: I will be gone

Hello, I have found a problem with using the glm function with a gamma family. I have a vector of data, assumed to be generated by a gamma distribution. The parameters of this gamma distribution are estimated in two ways (i) using the glm() function, (ii) "by hand", using the optim() function. I find that the -2*likelihood at the maximum found by (i) is substantially larger than that

Hello, I am currently working with the betareg package, which allows the fitting of a variable dispersion beta regression model (Simas et al. 2010, Computational Statistics & Data Analysis). I was wondering whether there is any package in R that allows me to fit variable dispersion parameters in the standard logistic regression model, that is to make the dispersion parameter contingent upon

Greetings, I'm posting this OT query here because of out very international membership! In the French sentence "Les taux de tirage sont calcul??s de mani??re ?? ce que la dispersion soit inf??rieure ?? 5 % dans chaque strate." it would seem intended that the "dispersion" is to be calculated in a specific way (unstated) -- otherwise, how to ensure that it shall be

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

Using R 1.8.1, and the negative binomial glm implemented in MASS, the default when using anova and a chi-square test is to divide the deviance by the estimated dispersion. Using my UNIX version of S-plus (v 3.4), and the same MASS functions, the deviances are *not* divided by the estimated dispersion. Firstly, I'm wondering if anyone can enlighten about the correct procedure (I thought

Hi, I'm programming in R and below is a summary of a generalized linear model: ************************************************** *** Call: glm(formula = offspring ~ degdays, family = quasi(link = "log", variance = "mu"), data = fecundity) Deviance Residuals: Min 1Q Median 3Q Max -0.76674 -0.29117 -0.09664 0.15668 1.00800 Coefficients: Estimate Std. Error t value

Hi all, First, a quick thank you for R; it's amazing. I am trying to fit models for a count dataset following the overdispersed logisitic regression approach outlined in Baggerly et al. (BMC Bioinformatics, 5:144; Annotated R code is given at the end of the paper) but R is returning an error with the data below. Any help in understanding or overcoming this obstacle is appreciated.