Part 1: 25 points
The previous time you tackled this section you had not covered how to deal with non-linearities in multiple regression and how to test regression assumptions and model fit in a rigorous way.
This is also a direct opportunity to apply the feedback you received on the first assessment.A team of researchers sets out on an ambitious project.
They want to test the effects of absoluteand relative wealth on well-being (life satisfaction/happiness).
To do so, they randomly sample 1000 sub-national regions out of all regions in the world. For each region, they measure satisfaction with life(well-being) on a 0-100 scale where 100 is the happiest.
To capture absolute wealth, they collect data on GDP/capita in each region in 1000 US dollars. Relative wealth is measured using the Gini-index. Itranges from 0-100, where 100 implies that all wealth is in the hands of one person and 0 implies perfectequality.
Questions
1. Formulate hypotheses for both wealth variables and justify them. Introduce the research question to the reader.
2. Regress life satisfaction on the Gini coefficientandGDP capita (life satisfaction is the dependent variable) and present your results in a publication-style regression table. Also run two bivariate models for each of your independent variables.
3. Compare you multiple regression model to the results from bivariate models. Discuss model fit.Can you run test to compare it?
4. Interpret your multiple regression model in statistical and in substantive terms.
5. Construct confidence intervals around the estimates of the coefficient for theGini Goefficientandthe coefficient forGDP/captiaand interpret them.
6. Re-examine the scatter plots, and address non-linearities. Try running a fourth model usingthe tools you have learned to address non-linearity
