search for: linkinv

...difference.criterion) stop (paste("POD.ceiling-POD.floor difference must be greater than ",difference.criterion," to discourage answer-shopping.", sep="")) linkfun <- function(mu) { mu <- qlogis( (mu - POD.floor)/(POD.ceiling - POD.floor) ) } linkinv <- function(eta) { eta <- POD.floor + (POD.ceiling - POD.floor)*plogis(eta) } mu.eta <- function(eta) { (POD.ceiling - POD.floor)*dlogis(eta)# derivitaive of mu with respect to eta } valideta <- function(eta) TRUE link <- paste("logit.FC("...

Dear all I am using NAMESPACE in my package but I would like the user to be able to overwrite four functions: own.linkfun, own.linkinv, own.mu.eta and own.valideta. These are used to defined "own" link functions. Is there any way of doing that without getting the when I am checking the package? This is what I am getting: make.link.gamlss : linkfun: no visible binding for global variable 'own.linkfun' make.l...

...al(cc, parent.frame())) 16 } 17 if (is.null(weights)) 18 weights <- rep(1, nobs) 19 if (is.null(offset)) 20 offset <- rep(0, nobs) 21 variance <- family$variance 22 dev.resids <- family$dev.resids 23 aic <- family$aic 24 linkinv <- family$linkinv 25 mu.eta <- family$mu.eta 26 if (!is.function(variance) || !is.function(linkinv)) 27 stop("illegal `family' argument") 28 valideta <- family$valideta 29 if (is.null(valideta)) 30 valideta <- function(eta) TRU...

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) }, stop(sQuote(link), " link not recognised")) structure(list(linkfun = linkfun, linkinv = linkinv, mu.eta = mu.eta, valideta...

.../library/stats/R/family.R 2005-03-04 04:40:03.000000000 -0500 +++ R-beta-fix/src/library/stats/R/family.R 2005-04-14 08:30:03.000000000 -0400 @@ -25,9 +25,9 @@ else if(!is.character(link) && !is.na(lambda <- as.numeric(link))) { linkfun <- function(mu) mu^lambda linkinv <- function(eta) - pmax(.Machine$double.eps, eta^(1/lambda)) + pmax(eta^(1/lambda), .Machine$double.eps) mu.eta <- function(eta) - pmax(.Machine$double.eps, (1/lambda) * eta^(1/lambda - 1)) + pmax((1/lambda) * eta^(1/lambda - 1), .Machine$d...

...L seems to work, but as I want to evaluate a suite of competing models, I'd like to use ML or REML estimation methods in order to end up with meaningful log-likelihoods. Here's the link function I use: logexp <- function(days = 1) { linkfun <- function(mu) qlogis(mu^(1/days)) linkinv <- function(eta) plogis(eta)^days mu.eta <- function(eta) days*.Call("logit_mu_eta", eta, PACKAGE = "stats")*plogis(eta)^(days-1) valideta <- function(eta) TRUE link <- paste("logexp(", days, ")", sep="") str...

...to this problem. I have appended the function i wrote. Add.haz<-function () { env <- new.env(parent = .GlobalEnv) assign(".nziek", nziek, envir = env) famname<-"Addhaza" link="addlink" linkfun<-function(mu) -log(1-mu) linkinv<-function(eta) 1-exp(-eta) variance<-function(mu) mu*(1-mu) validmu<-function(mu) all(mu > 0) && all(mu < 1) mu.eta<-function(mu) 1/(1-mu) dev.resids <- function(y, mu, wt) 2 * wt * (y * log(ifelse(y == 0, 1, y/mu)) + (1 - y) *...

...mu) qlogis(2*mu-1) half.logit.inv=function(eta) .5*plogis(eta)+.5 half.logit.deriv=function(eta) .5*(exp(eta/2)+exp(-eta/2))^-2 half.logit.inv.indicator=function(eta) TRUE half.logit.indicator=function(mu) mu>.5 & mu<1 link <- "half.logit" structure(list(linkfun = half.logit, linkinv = half.logit.inv, ??????????????????? mu.eta = half.logit.deriv, validmu = half.logit.indicator ,valideta = half.logit.inv.indicator, name = link), ?????????????? class = "link-glm") } # submit halflogit function to lmer rcNrPSrcNxPSxP<-lmer(Correct~-1+cNoise*f_Pos+f_Shape_con+f_Shape_...

...my understanding is that lmer does not currently allow a custom link. Therefore, I was investigating if other procedures for mixed models will allow a custom link function. here is the custom link function: logexp <- function(days = 1) { linkfun <- function(mu) qlogis(mu^(1/days)) linkinv <- function(eta) plogis(eta)^days mu.eta <- function(eta) days * plogis(eta)^(days-1) * .Call("logit_mu_eta", eta, PACKAGE = "stats") valideta <- function(eta) TRUE link <- paste("logexp(", days, ")", sep="") struct...

...low to illustrate what I'm hoping to scale up to a mixed model. Thanks very much for your time and thoughts. ----------------- library(MASS) logexp <- function(days = 1) ##Custom link function from Shaffer (2004) Auk 121:526-540. { linkfun <- function(mu) qlogis(mu^(1/days)) linkinv <- function(eta) plogis(eta)^days mu.eta <- function(eta) days * plogis(eta)^(days-1) * .Call("logit_mu_eta", eta, PACKAGE = "stats") valideta <- function(eta) TRUE link <- paste("logexp(", days, ")", sep="") struct...

...variancetemp), domain = NA)) initialize <- expression({ n <- rep.int(1, nobs) mustart <- y + 0.1 * (y == 0) }) aic <- function(y, n, mu, wt, dev) NA structure(list(family = "quasi", link = linktemp, linkfun = stats$linkfun, linkinv = stats$linkinv, variance = variance, dev.resids = dev.resids, aic = aic, mu.eta = stats$mu.eta, initialize = initialize, validmu = validmu, valideta = stats$valideta, varfun = variancetemp), class = "family") }

...d(seed) n <- length(fitted(fit)) theta <- model.matrix(fit$terms, data=fit$data) %*% coef(fit) ymat <- matrix(NA, nrow=n, ncol=nsim) for (s in 1:nsim) { if ("poisson"==fit$family$family) { ymat[,s] <- rpois(n, fit$family$linkinv(theta)) } else if ("binomial"==fit$family$family) { ymat[,s] <- rbinom(n, 1, fit$family$linkinv(theta)) } } as.data.frame(ymat) } # Eg: simulate.glm.hack(fit.poisson)[[1]] # [1] 1 6 4 7 5 9 7 3 3 4 2 4 3 7 4 4 7 2 4 6 simula...

...pred <- predict.lm(object, newdata, se.fit, scale = 1, type = ifelse(type == "link", "response", type), terms = terms, na.action = na.action) switch(type, response = {pred <- family(object)$linkinv(pred)}, link = , terms = ) } } else { ## summary.survreg has no ... argument. if(inherits(object, "survreg")) dispersion <- 1. if(is.null(dispersion) || dispersion == 0) dispersion <- summary(object, dispersion=dispersion)$dispersion...

I would like to do an R glm() with family = binomial(link="loglog") Right now, the cloglog link exists, which is nice when the data have a heavy tail to the left. I have the opposite case and the loglog link is what I need. Can someone suggest how to add the loglog link onto glm()? It would be lovely to have it there by default, and it certainly makes sense to have the two opposite

...else { pred <- predict.lm(object, newdata, se.fit, scale = 1, type = ifelse(type == "link", "response", type), terms = terms, na.action = na.action) switch(type, response = { pred <- family(object)$linkinv(pred) }, link = , terms = ) } } else { if (inherits(object, "survreg")) dispersion <- 1 if (is.null(dispersion) || dispersion == 0) dispersion <- summary(object, dispersion = dispersion)$dispersion residual.scale <- as.vector(sqrt(di...

...0) Y <- as.double(glmFit$y) mer <- .Call(mer_create, fl, Zt, X, Y, 0, nc, cnames) if (!is.null(start)) mer <- setOmega(mer, start) gVerb <- getOption("verbose") weights <- glmFit$prior.weights eta <- glmFit$linear.predictors linkinv <- quote(family$linkinv(eta)) mu.eta <- quote(family$mu.eta(eta)) mu <- family$linkinv(eta) variance <- quote(family$variance(mu)) dev.resids <- quote(family$dev.resids(Y, mu, weights)) LMEopt <- get("LMEoptimize<-") doLMEopt <- quote(LMEopt...

...ub.com/wch/r-source/blob/trunk/src/library/stats/man/family.Rd>]) it says: valid.mu: logical function. Returns ?TRUE? if a mean vector ?mu? is within the domain of ?variance?. valid.eta: logical function. Returns ?TRUE? if a linear predictor ?eta? is within the domain of ?linkinv?. However, the real names of these elements are "validmu" and "valideta" (no dots); try names(poisson()) or names(binomial()) and see ... Can someone on R-core fix this and save me the trouble of filing a bug report ... ? Or have I misunderstood something? cheers...

I am trying to get a prediction interval from a glm regression. With newdat being my set of values to be fitted, and glmreg the name of my regression, I am using the following code. predict(glmreg, newdat, se.fit = TRUE, interval = "confidence", level = 0.90) The problem is that I am only getting the standard error and the fitted value, not a prediction interval. Any help would be

...y=binomial("logit"),data=emrisk) #predictions for age 30 and 60 preds<-predict(model2d,data.frame(Age=c(30,60)),se.fit=TRUE) # prediction interval predsxx<-cbind(fit=preds$fit,lower=preds$fit-1.96*preds$se,upper=preds$fit+1.96*preds$se) #transformation model2dres<-family(model2d)$linkinv(predsxx) In my next step I want to know the confidence interval (CI) for the change in risk for the allergic event to occur for age 60 compared to age 30. The estimates from the model suggest a 80 pct. higher risk for age 60 compared to age 30. (100*model2dres[2]/model2dres[1]) But how should...

...he following: svec<-c(qlogis(mean(data1$scaled)),0,0,0) f<-betareg(scaled~-1 | expt_label + grouped_hpi, data=data1, link.phi="log", control=betareg.control(start=svec)) I understood that y~-1 could be used to give a fixed mean of 0.5 however I get the following error: Error in linkinv(x %*% beta + offset) : Argument eta must be a nonempty numeric vector I think I can work round this by using: svec2<-c(qlogis(mean(data1$scaled)),0,0,0,0,0) f2<-betareg(scaled ~ expt_label + grouped_hpi | expt_label + grouped_hpi, data=data1, + link.phi="log",con...