predictions(., hypothesis) arguments inconsistency bw python and R

[apoorvalal]

The accompanying python replication code to the JSS article does not work at the following two lines

marg.predictions(m, by="democracy", hypothesis="b2 = b1 * 2")
marg.predictions(
    m, by="democracy", hypothesis="b2 = b1 * 2", equivalence=[-0.2, 0.2]
).select("term", "estimate", "std_error", "p_value_equiv")

with NameError: name 'b2' is not defined.

A notebook demo-ing the correct python code either in the repo or the marginal effect zoo would be great.

[vincentarelbundock]

lol this is embarrassing.

Crazy week, but I'll look into it soon. Should be an easy fix.

Or maybe @artiom-matvei has time to take a look.

[vincentarelbundock]

Might be a b0 vs b1 issue. I think we decided it was necessary to make a breaking change so that Python used zero indexing.

[artiom-matvei]

@vincentarelbundock is there a way for me to update the replication code?

[pvsundar]

From @drsundar.bsky.social

#corrected code chunk:
#Merely change b2 to b1; and b1 to b0.

avg_predictions(m,
by = "democracy",
hypothesis = "pairwise")

predictions(m,
by = "democracy",
hypothesis = "b1 = b0 * 2")

predictions(m,
by = "democracy",
hypothesis = "b1 = b0 * 2",
equivalence = [-.2, .2])
.select("term", "estimate", "std_error", "p_value_equiv")

[vincentarelbundock]

Thanks everyone for confirming that the index of hypotheses in Python must start at b0. This is a type in the Python replication package.

I will collect a list of errors and typos found in the JSS here: vincentarelbundock/marginaleffects#1293

After a while, I'll try to update the package at JSS, but I'm not sure if they will let me.