rqgam_check: Randomised quantile residuals check plot for GAMs/DSMs

In dsm: Density Surface Modelling of Distance Sampling Data

Randomised quantile residuals check plot for GAMs/DSMs

Description

Reproduces the "Resids vs. linear pred" plot from gam.check but using randomised quantile residuals, a la Dunn and Smyth (1996). Checks for heteroskedasticity as as usual, looking for "funnel"-type structures in the points, which is much easier with randomised quantile residuals than with deviance residuals, when your model uses a count distribution as the response.

Usage

rqgam_check(gam.obj, ...)

Arguments

gam.obj	a gam, glm or dsm object.
...	arguments passed on to all plotting functions

Details

Note that this function only works with negative binomial and Tweedie response distributions.

Earlier versions of this function produced the full gam.check output, but this was confusing as only one of the plots was really useful. Checks of k are not computed, these need to be done using gam.check.

Value

just plots!

Author(s)

Based on code by Natalie Kelly, bugs added by Dave Miller

Examples

library(Distance)
library(dsm)
library(tweedie)

# load the Gulf of Mexico dolphin data (see ?mexdolphins)
data(mexdolphins)

# fit a detection function and look at the summary
hr.model <- ds(distdata, truncation=6000,
               key = "hr", adjustment = NULL)

# fit a simple smooth of x and y with a Tweedie response with estimated
#  p parameter
mod1 <- dsm(count~s(x, y), hr.model, segdata, obsdata, family=tw())
rqgam_check(mod1)

dsm documentation built on Aug. 21, 2022, 1:07 a.m.