would be good to replace the grid value calculation in pdp_calc_utils.py line 237:
value_grids = np.percentile(feature_values, percentile_grids)
with
value_grids = np.nanpercentile(feature_values, percentile_grids)
This will avoid returning an array with nans only if the amount of nans is high in the dataset.
The latest version, v0.3.0, has been released. We recommend updating to this version and checking if it resolves the issue you're facing. Your feedback is appreciated. Thank you for your patience and support.
Hello,
would be good to replace the grid value calculation in pdp_calc_utils.py line 237:
value_grids = np.percentile(feature_values, percentile_grids)
withvalue_grids = np.nanpercentile(feature_values, percentile_grids)
This will avoid returning an array with nans only if the amount of nans is high in the dataset.