Getting negative values for test_calibration(tau.forest)

grf-labs / grf

Generalized Random Forests

https://grf-labs.github.io/grf/

GNU General Public License v3.0

938 stars 250 forks source link

Getting negative values for test_calibration(tau.forest) #1405

Closed YongzhWang closed 3 months ago

YongzhWang commented 3 months ago

Hi,

I am trying to fit a casual forest but get approx. -1 estimates for mean.forest.prediction. I wonder if it is common to get negative values for test_calibration(tau.forest). Does this just signal that the prediction is terrible?

Best, Andrew

erikcs commented 3 months ago

Yes. You might find the RATE easier to interpret for this kind of exercise.