Plotting/graph question

[Willem Vanlaer]

Hello,

I am currently learning how to work with Generalized Additive Models (GAMs).
I found the following exercise in an old Stata Technical Bulletin:


Now, producing the model is pretty straightforward, but I was wondering how to get this graph.
I have been searching a bit online for how to do this, but most (introductory) manuals simply explain how to produce a straight "trendline", not the ones shown above.
I have very limited experience in visualizing my results.

Thank you for your help,

Willem

[Sergiy Radyakin]

Willem, once you have the model, you just plot the fitted values. Best, Sergiy

clear all
sysuse auto
logit foreign mpg
predict hat
label variable hat "Predicted probability of foreign"
twoway scatter foreign mpg || line hat mpg,sort

[Willem Vanlaer]

Willem, once you have the model, you just plot the fitted values. Best, Sergiy

clear all
sysuse auto
logit foreign mpg
predict hat
label variable hat "Predicted probability of foreign"
twoway scatter foreign mpg || line hat mpg,sort

[ATTACH=CONFIG]n166029[/ATTACH]

Hey Sergiy,

Thank you for your reply.
I have tried this, but I get the following error message after having run the predict command: "last estimates not found".

My code is:

gam growth debt, df(5)
predict predgrowth
label variable predgrowth "Predicted growth"
twoway scatter growth debt || line predgrowth debt,sort

What am I doing wrong?

[Nick Cox]

Not reading the help carefully....

gam creates a new variable GAM_mu containing the fitted values on the scale of the response variable.

gam (STB-42) is a rather old and non-standard program. Evidently it doesn't support predict, as is now fairly standard for programs of its kind. But the help explains that it leaves predicted values in memory in a new variable, which is what you need here.

[Willem Vanlaer]

Not reading the help carefully....



gam (STB-42) is a rather old and non-standard program. Evidently it doesn't support predict, as is now fairly standard for programs of its kind. But the help explains that it leaves predicted values in memory in a new variable, which is what you need here.

Thanks. It works now!