cat_template breaks with labels option included

[donbowen]

Hi @s3alfisc @leostimpfle

import numpy as np
import pyfixest as pf
data = pf.get_data()
data['job'] = np.random.choice(["Managerial", "Admin", "Blue collar"], size=len(data), p=[1/3, 1/3, 1/3])

fit = pf.feols("Y ~ X1 + X2 + job", data = data)

# works:
pf.etable([fit], cat_template = "weeee::{value}")

# fails
labels = {
    "Y": "Wage",
    "Y2": "Wealth",
    "X1": "Age",
    "X2": "Years of Schooling",
    "f1": "Industry",
    "f2": "Year",
}

pf.etable([fit_i], cat_template = "sadface::{value}", labels=labels)

# this, from the docs, doesn't output whats in the doc webpage either
labels['job']="Job Family"
fit7 = pf.feols("Y ~ X1 + X2 + job", data = data)
pf.etable([fit7], labels=labels, cat_template = "{variable}::{value}")

[leostimpfle]

Hi @donbowen, I can't immediately see what's not working based on the code you've shared. Could you please elaborate? (Note that I've used fit across all pf.etable calls because it's not clear what fit_i refers to.)

[donbowen]

I accidentally omitted a line above it.

import numpy as np
import pyfixest as pf
data = pf.get_data()
data['job'] = np.random.choice(["Managerial", "Admin", "Blue collar"], size=len(data), p=[1/3, 1/3, 1/3])

fit = pf.feols("Y ~ X1 + X2 + job", data = data)

# works:
pf.etable([fit], cat_template = "weeee::{value}")

# fails
labels = {
    "Y": "Wage",
    "Y2": "Wealth",
    "X1": "Age",
    "X2": "Years of Schooling",
    "f1": "Industry",
    "f2": "Year",
}

fit_i = pf.feols("Y ~ X1 + X2 + i(job)", data = data)
pf.etable([fit_i], cat_template = "sadface::{value}", labels=labels)

# this, from the docs, doesn't output whats in the doc webpage either
labels['job']="Job Family"
fit7 = pf.feols("Y ~ X1 + X2 + job", data = data)
pf.etable([fit7], labels=labels, cat_template = "{variable}::{value}")

So, there are 3 tables here, and the first is like I expect, but the fit_i one isn't, nor is fit7.

The behavior changes when they are combined.

# with this - fit_i and fit7 suddenly do what cat_template says, though they aren't treated as the same X vars (ie duplicated rows for blue collar coefs)
pf.etable([fit, fit_i, fit7], labels=labels, cat_template = "{variable}::{value}")

[felixschmitz]

Hi @donbowen, thanks for bringing this up! Could you please re-iterate for me, what does not work as expected?
I could reproduce the behavior of combining multiple instances of ETable and having multiple coefficients with the same labels ("i.e. duplicated rows for blue collar coefs").

[donbowen]

fit "Y ~ X1 + X2 + job"

pf.etable([fit], cat_template = "weeee::{value}")

The template works and you see "weeee" in the variable labels:

fit_i: "Y ~ X1 + X2 + i(job)"

pf.etable([fit_i], cat_template = "sadface::{value}", labels=labels)

The template fails and you do not see "sadface" in the variable labels:

The only difference between those two is that fit_i Is an estimation using i(job) instead of job in the formula.

fit7: "Y ~ X1 + X2 + job", as in fit, but labels doesn't work as expected

pf.etable([fit7], labels=labels, cat_template = "{variable}::{value}")

Fit7 here is the same as the first one above that "works" but labels['job']="Job Family" doesn't work. And it looks different from what the docs show for the same thing.

combined

job and i(job) equations aren't combined, labels.

[felixschmitz]

Thanks, @donbowen for elaborating! I am unable to reproduce your rendering errors on fit_i and fit7 with pyfixest version 0.30.2 and maketables version 0.1.9. Could you please tell me which version of pyfixest and maketables you are using?
Regarding your combination of multiple models into one table, in my opinion this is the expected behavior. Your models fit and fit7 are equivalent in formula and data, while your model fit_i has a different formula. The returned coefficients are thus different in type (as your last screenshot also indicates e.g. "job[T.Mangerial]" and "C(job)[T.Managerial]".

Hope this helps, but please let me know, if the problem persists for you!