Can one extract model fit parameters after a ggplot stat_smooth call?

Question

Using stat_smooth, I can fit models to data. E.g.

g=ggplot(tips,aes(x=tip,y=as.numeric(unclass(factor(tips$sex))-1))) +facet_grid(time~.) 
g=g+ stat_summary(fun.y=mean,geom="point") 
g=g+ stat_smooth(method="glm", family="binomial")

I would like to know the coefficients of the glm binomial fits. I could re-do the fit with dlply and get the coefficients with ldply, but I'd like to avoid such duplication.

Calling str(g) reveals the hierarchy of objects that ggplot2 creates, perhaps there's some way to get to the coefficients through that?

P.S. is there a better way to change sex to a number besides the as.numeric etc. method above? — Alex Holcombe, Apr 19 '10 at 11:03

score 3 · Accepted Answer · edited Oct 01 '19 at 06:15

3

No, because the models are only created when the plot is rendered. However, it's usually pretty easy to do it yourself with plyr.

Why do you want to convert sex to a number? Using as.numeric should be enough by itself, but if you're going to do the subtraction in the model you'll need to surround it with I().

edited Oct 01 '19 at 06:15

NelsonGon

13,015
7
27
57

answered Apr 19 '10 at 12:30

hadley

102,019
32
183
245

3

Is there a chance we could see how this is done with plyr? I have the same question, but no idea where to start. Thanks! – Atticus29 Jul 23 '12 at 03:20

Can one extract model fit parameters after a ggplot stat_smooth call?

1 Answers1

Linked