the predictors of voting for Barack Obama in the 2012 US presidential election.
Data Preparation and Description:
Prepare Data:
Create a new variable for religion with categories: Catholic, Protestant, Other, and None (keeping NAs). Generate a new variable indicating if either parent has a bachelor or graduate degree. Filter data for non-missing values in the religion, education, obama, age, and sex variables.
Describe Sample:
Calculate descriptive statistics for the new religion and education variables, age, sex, and obama. Present shares of observations for categorical variables and mean/standard deviations for continuous ones. Perform analysis for the full sample and separately for Obama and non-Obama voters.
Brief Discussion:
Discuss the composition of the sample across religion, education, age, sex, and Obama voting status.
Model Estimation and Odds Ratios:
Logistic Regression Models:
a) Religion variable only.
b) Religion and parental education variables.
c) Religion, parental education, sex, and age variables.
Presentation:
Show odds ratios with confidence intervals for all three models in a single table. Include key model statistics (N, log-likelihood) at the bottom of the table. Present odds ratios from Model c in a graph without the intercept, with confidence intervals.
Discussion:
Interpret results of logistic regression models, considering the impact of religion, education, sex, and age on voting for Obama.
Predicted Probabilities:
Graphical Visualization:
Plot predicted probabilities of voting for Obama at different ages for each religion category.Base predictions on Model c, assuming a female with at least one college-educated parent.
Discussion:
Interpret how predicted probabilities vary across age and religion categories, considering parental education.
Model Fit:
Nested Models:
Explain nesting relationships between models a, b, and c. Perform likelihood ratio tests comparing model b to model a and model c to model b.
Fit Statistics:
Calculate Nagelkerke’s pseudo-R2 and share of correctly predicted observations for all three models. Present fit statistics in a well-formatted table.
Discussion:
Interpret results of model fit comparisons, assessing the improvement in model fit with added variables.