`avg_comparisons(model, hypothesis="revpairwise")` does not work in Python without the `by` argument while it works in R
Opened this issue · 1 comments
artiom-matvei commented
R code
> library(marginaleffects)
> dat = read.csv("https://vincentarelbundock.github.io/Rdatasets/csv/palmerpenguins/penguins.csv")
> mod = glm(body_mass_g ~ flipper_length_mm * species * bill_length_mm + island, data = dat)
> c = avg_comparisons(mod, hypothesis = "revpairwise")R output
Term Estimate Std. Error z Pr(>|z|) S 2.5 % 97.5 %
(flipper_length_mm, mean(+1)) - (bill_length_mm, mean(+1)) -37.3 9.52 -3.921 < 0.001 13.5 -56 -18.7
(island, mean(Dream) - mean(Biscoe)) - (bill_length_mm, mean(+1)) -75.0 67.71 -1.108 0.26793 1.9 -208 57.7
(island, mean(Torgersen) - mean(Biscoe)) - (bill_length_mm, mean(+1)) -122.5 69.72 -1.757 0.07896 3.7 -259 14.2
(species, mean(Chinstrap) - mean(Adelie)) - (bill_length_mm, mean(+1)) -578.3 201.74 -2.867 0.00415 7.9 -974 -182.9
(species, mean(Gentoo) - mean(Adelie)) - (bill_length_mm, mean(+1)) -430.5 253.67 -1.697 0.08970 3.5 -928 66.7
(island, mean(Dream) - mean(Biscoe)) - (flipper_length_mm, mean(+1)) -37.7 68.00 -0.554 0.57929 0.8 -171 95.6
(island, mean(Torgersen) - mean(Biscoe)) - (flipper_length_mm, mean(+1)) -85.2 69.97 -1.217 0.22357 2.2 -222 52.0
(species, mean(Chinstrap) - mean(Adelie)) - (flipper_length_mm, mean(+1)) -541.0 199.72 -2.709 0.00675 7.2 -932 -149.6
(species, mean(Gentoo) - mean(Adelie)) - (flipper_length_mm, mean(+1)) -393.2 254.07 -1.548 0.12174 3.0 -891 104.8
(island, mean(Torgersen) - mean(Biscoe)) - (island, mean(Dream) - mean(Biscoe)) -47.5 65.08 -0.729 0.46577 1.1 -175 80.1
(species, mean(Chinstrap) - mean(Adelie)) - (island, mean(Dream) - mean(Biscoe)) -503.3 221.75 -2.270 0.02322 5.4 -938 -68.7
(species, mean(Gentoo) - mean(Adelie)) - (island, mean(Dream) - mean(Biscoe)) -355.5 253.04 -1.405 0.16008 2.6 -851 140.5
(species, mean(Chinstrap) - mean(Adelie)) - (island, mean(Torgersen) - mean(Biscoe)) -455.9 209.09 -2.180 0.02925 5.1 -866 -46.0
(species, mean(Gentoo) - mean(Adelie)) - (island, mean(Torgersen) - mean(Biscoe)) -308.0 250.11 -1.231 0.21814 2.2 -798 182.2
(species, mean(Gentoo) - mean(Adelie)) - (species, mean(Chinstrap) - mean(Adelie)) 147.8 219.56 0.673 0.50071 1.0 -282 578.2
Type: response
Columns: term, estimate, std.error, statistic, p.value, s.value, conf.low, conf.high
Python code
import polars as pl
import statsmodels.formula.api as smf
from marginaleffects import *
penguins = pl.read_csv(
"https://vincentarelbundock.github.io/Rdatasets/csv/palmerpenguins/penguins.csv",
null_values="NA",
).drop_nulls()
mod = smf.ols(
"body_mass_g ~ flipper_length_mm * species * bill_length_mm + island",
penguins.to_pandas(),
).fit()
p = avg_comparisons(penguins, hypothesis='revpairwise')Python output
# truncated stacktrace
File c:\Users\IBM\Projects\Vincent A.B\pymarginaleffects\marginaleffects\sanitize_model.py:24, in sanitize_model(model)
[21](file:///C:/Users/IBM/Projects/Vincent%20A.B/pymarginaleffects/marginaleffects/sanitize_model.py:21) try:
[22](file:///C:/Users/IBM/Projects/Vincent%20A.B/pymarginaleffects/marginaleffects/sanitize_model.py:22) import pyfixest # noqa
---> [24](file:///C:/Users/IBM/Projects/Vincent%20A.B/pymarginaleffects/marginaleffects/sanitize_model.py:24) return ModelPyfixest(model)
[25](file:///C:/Users/IBM/Projects/Vincent%20A.B/pymarginaleffects/marginaleffects/sanitize_model.py:25) except ImportError:
[26](file:///C:/Users/IBM/Projects/Vincent%20A.B/pymarginaleffects/marginaleffects/sanitize_model.py:26) pass
File c:\Users\IBM\Projects\Vincent A.B\pymarginaleffects\marginaleffects\model_pyfixest.py:10, in ModelPyfixest.__init__(self, model)
[9](file:///C:/Users/IBM/Projects/Vincent%20A.B/pymarginaleffects/marginaleffects/model_pyfixest.py:9) def __init__(self, model):
---> [10](file:///C:/Users/IBM/Projects/Vincent%20A.B/pymarginaleffects/marginaleffects/model_pyfixest.py:10) super().__init__(model)
[11](file:///C:/Users/IBM/Projects/Vincent%20A.B/pymarginaleffects/marginaleffects/model_pyfixest.py:11) if hasattr(self.model, "_fixef"):
[12](file:///C:/Users/IBM/Projects/Vincent%20A.B/pymarginaleffects/marginaleffects/model_pyfixest.py:12) if self.model._fixef is not None:
File c:\Users\IBM\Projects\Vincent A.B\pymarginaleffects\marginaleffects\model_abstract.py:10, in ModelAbstract.__init__(self, model)
[8](file:///C:/Users/IBM/Projects/Vincent%20A.B/pymarginaleffects/marginaleffects/model_abstract.py:8) def __init__(self, model):
[9](file:///C:/Users/IBM/Projects/Vincent%20A.B/pymarginaleffects/marginaleffects/model_abstract.py:9) self.model = model
---> [10](file:///C:/Users/IBM/Projects/Vincent%20A.B/pymarginaleffects/marginaleffects/model_abstract.py:10) self.validate_coef()
[11](file:///C:/Users/IBM/Projects/Vincent%20A.B/pymarginaleffects/marginaleffects/model_abstract.py:11) self.validate_modeldata()
[12](file:///C:/Users/IBM/Projects/Vincent%20A.B/pymarginaleffects/marginaleffects/model_abstract.py:12) self.validate_response_name()
File c:\Users\IBM\Projects\Vincent A.B\pymarginaleffects\marginaleffects\model_abstract.py:17, in ModelAbstract.validate_coef(self)
[16](file:///C:/Users/IBM/Projects/Vincent%20A.B/pymarginaleffects/marginaleffects/model_abstract.py:16) def validate_coef(self):
---> [17](file:///C:/Users/IBM/Projects/Vincent%20A.B/pymarginaleffects/marginaleffects/model_abstract.py:17) coef = self.get_coef()
[18](file:///C:/Users/IBM/Projects/Vincent%20A.B/pymarginaleffects/marginaleffects/model_abstract.py:18) if not isinstance(coef, np.ndarray):
[19](file:///C:/Users/IBM/Projects/Vincent%20A.B/pymarginaleffects/marginaleffects/model_abstract.py:19) raise ValueError("coef must be a numpy array")
File c:\Users\IBM\Projects\Vincent A.B\pymarginaleffects\marginaleffects\model_pyfixest.py:18, in ModelPyfixest.get_coef(self)
[17](file:///C:/Users/IBM/Projects/Vincent%20A.B/pymarginaleffects/marginaleffects/model_pyfixest.py:17) def get_coef(self):
---> [18](file:///C:/Users/IBM/Projects/Vincent%20A.B/pymarginaleffects/marginaleffects/model_pyfixest.py:18) return np.array(self.model._beta_hat)
AttributeError: 'DataFrame' object has no attribute '_beta_hat'
vincentarelbundock commented
Thanks for working on this. Good issue to raise.