CIGLM {BSagri} | R Documentation |
Computes confidence intervals from the output of a glm, by calling to glht(multcomp).
CIGLM(x, conf.level = 0.95, method = c("Raw", "Adj", "Bonf"))
x |
a object of class "glm" , "gamlss" , or "glm.nb" as can be obtained by calling to function glm , function gamlss in package gamlss , or function glm.nb in package MASS |
conf.level |
confidence level, a single numeric value between 0.5 and 1 |
method |
a single character string, with "Raw" for unadjusted intervals, "Adj" for multiplicity adjusted intervals taking the correlations into account, or "Bonf" for a Bonferroni adjustment |
This is just a wrapper to confinf.glht
of package multcomp
.
Note that except for the simple general linear model with assumption of Gaussian response, the resulting intervals are exact intervals. In other cases, the methods are only asymptotically correct, hence might give misleading results for small sample sizes!
An object of class "confint.glht"
confint.glht
in package multcomp
for the function that is used internally,
UnlogCI
for a simple function to bring confidence intervals back to the original scales
when there is a log or logit link, with appropriate naming.
data(Diptera) library(multcomp) modelfit <- glm(Ges ~ Treatment, data=Diptera, family=quasipoisson) comps <- glht(modelfit, mcp(Treatment="Tukey")) CIs<-CIGLM(comps, method="Raw") CIs CIsAdj<-CIGLM(comps, method="Adj") CIsAdj CIsBonf<-CIGLM(comps, method="Bonf") CIsBonf library(gamlss) modelfit2 <- gamlss(Ges ~ Treatment, data=Diptera, family=NBI) comps2 <- glht(modelfit2, mcp(Treatment="Tukey")) CIs2<-CIGLM(comps2, method="Raw") CIs2 CIsAdj2<-CIGLM(comps2, method="Adj") CIsAdj2 CIsBonf2<-CIGLM(comps2, method="Bonf") CIsBonf2