-
Notifications
You must be signed in to change notification settings - Fork 12
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Fast marginal effect plots (i.e., "poor man's PDPs") #91
Comments
Thanks you so much for the fantastic work on this package. It's been an absolute life-saver. Had a couple of questions for you about this:
Is the difference between what this does and what partial does is that, in the partial case, all the covariates keep their real values (instead of being fixed at their median), so you're getting predictions averaged over every real value of the other variables, rather than averaged over the median values? If so, where does the speed-up come from?
|
Thanks @DeFilippis, glad you've found the package useful. And I've been meaning to come back to this. Responses to your questions below:
The code above is nothing more than a way to trick the |
Perfect! Thanks so much for the thorough and speedy replies. Really appreciate it. |
Rather than averaging over the entire training set, you can fix all other features at their median (numeric features) or most frequent (categorical features) value (if there are no interaction effects, these plots will be parallel to the corresponding PDPs). This is similar in spirit to the excellent plotmo package:
The text was updated successfully, but these errors were encountered: