l_fitDens: Adding density strip of fitted effect

In mgcViz: Visualisations for Generalized Additive Models

Adding density strip of fitted effect

Description

This layer adds a conditional posterior density strip to 1D smooth effects plots. With the default colour scale, the opacity is proportional to the conditional density of the fitted effects, under the usual Gaussian approximation the posterior.

Usage

l_fitDens(n = 50, level = 0.95, trans = identity, ...)

Arguments

n	sqrt of the number of grid points used to compute the effect plot.
level	confidence level. By default the conditional density of the fit will be plotted between the Gaussian quantiles 0.025 and 0.975, hence the level determines the width of the y-axis.
trans	monotonic function to be applied to the density of the fit, which determines colour of the plot. Monotonicity is not checked.
...	further arguments to be passed to ggplot2::geom_raster.

Details

See Bowman (2018) for explanations about the advantages of density strips, relative to plots including the mean fit + confidence intervals.

Value

An object of class gamLayer.

References

Bowman, D. W (2018). Graphics for uncertainty. Journal of the Royal Statistical Society: Series A.

Examples

library(mgcViz)
set.seed(44)
dat <- gamSim(1,n=400,dist="normal",scale=2)
b <- gamV(y~s(x0)+x1+s(x2)+s(x3),data=dat)

plot(sm(b, 1)) + l_fitDens() + l_fitLine()
plot(pterm(b, 1)) + l_fitDens(trans = function(x) x^0.25) + l_fitLine()
               

mgcViz documentation built on Oct. 6, 2023, 5:09 p.m.