5  Regression and Other Stories: National election study

Farhan Reynaldo

Nov 16th, 2021

from pathlib import Path

import arviz as az
import bambi as bmb
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd

DATA_PATH = Path("../../data")

5.0.1 Load packages

rng = np.random.default_rng(seed=12)
az.style.use("arviz-white")

5.0.2 Load data

nes = pd.read_csv(DATA_PATH / "nes.txt", delimiter=" ")
nes.head()
year resid weight1 weight2 weight3 age gender race educ1 urban ... parent_party white year_new income_new age_new vote.1 age_discrete race_adj dvote rvote
536 1952 1 1.0 1.0 1.0 25 2 1 2 2.0 ... 2.0 1 1 1 -2.052455 1.0 1 1.0 0.0 1.0
537 1952 2 1.0 1.0 1.0 33 2 1 1 2.0 ... 0.0 1 1 1 -1.252455 1.0 2 1.0 1.0 0.0
538 1952 3 1.0 1.0 1.0 26 2 1 2 2.0 ... -2.0 1 1 0 -1.952455 1.0 1 1.0 0.0 1.0
539 1952 4 1.0 1.0 1.0 63 1 1 2 2.0 ... NaN 1 1 0 1.747545 1.0 3 1.0 0.0 1.0
540 1952 5 1.0 1.0 1.0 66 2 1 2 2.0 ... -2.0 1 1 -2 2.047545 1.0 4 1.0 0.0 1.0

5 rows × 70 columns

5.0.3 Partyid model to illustrate repeated model use (secret weapon)

def regress_year(year, data=nes):
    this_year = data[data["year"] == year]
    model = bmb.Model(
        "partyid7 ~ real_ideo + race_adj + C(age_discrete) + educ1 + female + income",
        data=this_year,
        dropna=True,
    )
    idata = model.fit()
    stats = az.summary(idata, kind="stats")
    return (year, stats)


years = np.arange(1972, 2004, 4)
summary = [regress_year(year) for year in years]
100.00% [8000/8000 00:04<00:00 Sampling 4 chains, 0 divergences]
100.00% [8000/8000 00:04<00:00 Sampling 4 chains, 0 divergences]
100.00% [8000/8000 00:04<00:00 Sampling 4 chains, 0 divergences]
100.00% [8000/8000 00:04<00:00 Sampling 4 chains, 0 divergences]
100.00% [8000/8000 00:04<00:00 Sampling 4 chains, 0 divergences]
100.00% [8000/8000 00:04<00:00 Sampling 4 chains, 0 divergences]
100.00% [8000/8000 00:05<00:00 Sampling 4 chains, 0 divergences]
100.00% [8000/8000 08:02<00:00 Sampling 4 chains, 0 divergences]

5.0.4 plot

coefs = [
    "Intercept",
    "real_ideo",
    "race_adj",
    "C(age_discrete)[2]",
    "C(age_discrete)[3]",
    "C(age_discrete)[4]",
    "educ1",
    "female",
    "income",
]

coef_names = [
    "Intercept",
    "Ideology",
    "Black",
    "Age_30_44",
    "Age_45_64",
    "Age_65_up",
    "Education",
    "female",
    "income",
]

fig, axes = plt.subplots(2, 5, figsize=(15, 9))

for ax, coef in zip(axes.flat, coefs):
    coef_each_year = [summary[i][1].loc[coef, "mean"] for i in range(len(summary))]
    coef_sd_each_year = [summary[i][1].loc[coef, "sd"] for i in range(len(summary))]
    ax.scatter(years, coef_each_year, color="black")
    ax.errorbar(
        years, coef_each_year, yerr=coef_sd_each_year, fmt="none", color="black"
    )
    ax.axhline(0, linestyle="--", color="gray")
    ax.set_title(coef_names[coefs.index(coef)])

axes.flat[-1].remove()

4  Regression and Other Stories: Congress