Plotting a model that uses a smoother with knots specified by a variable

[DanielReedOcean]

First of all, kudos on an awesome package! I'm having a bit of a problem when specifying the number of knots in a smoother with a variable and then plotting the results with visreg. For example, the following seems to think kval, the variable for the number of knots used by the smoother, is a predictor and creates a plot for it:

# Load data
data("mtcars")

# Number of knots in the smoother function
kval <- 3

# Create a formula that looks like this: mpg ~ s(hp, k = kval)
formula.tmp <- reformulate(response = "mpg", termlabels = "s(hp, k = kval)")

# Run GAM
gam.res <- gam(formula.tmp, data = mtcars)

# Load library
library(visreg)

# Plot results 
visreg(gam.res)
##### This plots for two predictors: hp and kval. #####

If I hardcode the number of knots in the formula, this doesn't happen even though the models are essentially exactly the same.

# Create a formula that looks like this: mpg ~ s(hp, k = 3)
formula.tmp <- reformulate(response = "mpg", termlabels = "s(hp, k = 3)")

# Run GAM
gam.res <- gam(formula.tmp, data = mtcars)

# Plot results 
visreg(gam.res)
##### This plots for one predictor: hp. #####

[pbreheny]

Daniel,

Thanks for bringing this up...it didn't really rise high on my radar since you can always just request the specific plot you want with

visreg(gam.res, "hp")

But still, I agree, visreg was not doing the correct thing here -- fixed now, though.