You will choose a state and use its data file to predict household income (HINCP) using other variables in the data set using mulitple linear regression. Remember that there are two files whose variables are described in the data dictionary: household variables that appear in this file, and individual variables that appear in the file we used for the previous project. Since you’re just using the household file, remember you won’t be able to use any individual variables (e.g. age, sex, commute time, education, etc).
Your final model should have a reasonable r squared and an overall “significant” p-value. Do not use FINCP as a predictor! FINCP is family income, and saying that we can predict a household’s income if we know the family’s income is very nearly a circular argument. For the vast majority of Americans, family income and household income are the same thing. Similarly, do not use any predictor variable (like OCPIP) that is calculated using household income. If you already know household income (which is used to calcuate OCPIP), then it should be pretty darn easy to predict household income! If any of your predictors have large p-values, be sure to justify why you are including them. Every student in the group should contribute to and comment on the body of the report, even if grammar and graph details are left to individual group members.
To really impress, make a prediction for a particular household with a given set of predictor variables.
Deliverable #2: Upload your report to Canvas by the deadline.
5 points: Model, including XLStat output including ANOVA and coefficient tables (XLStat calls that a “Model Parameters” table- don’t be confused by the “standardized coefficients” table) but NOT any automated “explanation” such as StatGraphic’s “The StatAdvisor”.
5 points: 1-4 complete sentences describing your model
5 points: Graph. Remember your principles of data viz!
5 points: Comment on outliers, patterns [this exploration should also include a scatterplot matrix for your predictor variables: pairs() in R or plot>multivariate visualization>scatterplot matrix in statgraphics or Visualizing Data>Scatterplot, then all your variables in both x and y in XLStat]
5 points: R squared better than 25% (partial credit for getting close)
5 points: Describe each independent variable (what is it?). If your final model included less than 4 predictors, describe other variables you considered but did not include. Describe a minimum of four variables.
5 points: For each independent variable, speculate on its sign and value (why is it positive/negative? is it “big” or “small” in its effect?)
