Myujth

I'm trying to draw in a single plot crude and adjusted GAM models using library visreg:

# Create DF
set.seed(123)
x1 = rnorm(2000) 
z = 1 + 3*x1 + 3*exp(x1) 
pr = 1/(1+exp(-z)) 
y = rbinom(2000,1,pr)
df = data.frame(y=y,x1=x1, x2=exp(x1)*z)

# Fitting GAMs
library(mgcv)
crude <- gam(y ~ s(x1), family=binomial(link=logit), data=df)
adj <- gam(y ~ s(x1) + s(x2), family=binomial(link=logit), data=df)

# Plot results using 'visreg'
library(visreg)
p.crude <- visreg(crude, scale='response', "x1", line.par = list(col = 'red'), gg=TRUE) + theme_bw()
p.adj <- visreg(adj, scale='response', "x1", gg=TRUE) + theme_bw()

Using gridExtra I can produce a two columns plot, however I would have a single plot which overlays the two model plots.

You can use the plot=FALSE parameter to get the data without the plots:

p.crude <- visreg(crude, scale='response', "x1", line.par = list(col = 'red'), plot=FALSE)
p.adj <- visreg(adj, scale='response', "x1", plot = FALSE)

And, then re-create it by hand:

dplyr::bind_rows(
 dplyt::mutate(p.crude$fit, plt = "crude"),
 dplyr::mutate(p.adj$fit, plt = "adj")
) -> fits

ggplot() +
 geom_ribbon(
 data = fits, 
 aes(x1, ymin=visregLwr, ymax=visregUpr, group=plt), fill="gray90"
 ) +
 geom_line(data = fits, aes(x1, visregFit, group=plt, color=plt)) +
 theme_bw()

https://github.com/pbreheny/visreg/blob/master/R/ggFactorPlot.R has all the other computations and geoms/aesthetics you can use in the recreation.