Question

I am using vis.gam() to plot the results from the following model:

gam.5 <- gam(mortality.under.2~s(maternal_age)+ s(birth_year) + 
          te(birth_year,maternal_age) + wealth + sex + 
         residence+ maternal_educ 
          ,data=colombia1,family="binomial") 

The code to make the plot is the following:

 vis.gam(gam.5,view=c("maternal_age","birth_year"),
         plot.type="contour",type="response",
         cond=list(wealth="Lowest quintile",sex="Female",
         residence="Rural",maternal_educ="Primary"))

Unfortunately, it produces predict values outside the range of the data for the variables ‘maternal_age’ and ‘birth_year’. Any suggestion on how to constrain the range of the predicted values in the plot for the variables in view?

Thanks a bunch,

Antonio Pedro.

Answer 1

Looking at the code it appears the graphics paradigm is base graphics (specifically the persp function) and that there is trailing dots argument, so why not add:

..., `xlim=range(columbia1(maternal_age), ylim=range(columbia1$birth_year)`