Data also found in R:

library(wooldridge)

data("k401ksubs")

Consider the data set "k401ksubs.csv" posted with this assignment. It includes information on 9275 individuals with the following covariates, where the dependent variable is pira, equal to 1 if the subject has an IRA.

e401k: =1 if eligible for 401(k)

inc: annual income, $1000s

marr: =1 if married

male: =1 if male respondent

age: in years

fsize: family size

nettfa: net total fin. assets, $1000

p401k: =1 if participate in 401(k)

pira: = 1 if have IRA (Individual Retirement Account)

incsq: income squared

agesq: age squared

Question 3

Create 2 Logistic Regression Models

(1) using all variables (Model 1),

(2) using variables you deem important (Model 2)

a. In Model 1 interpret the impact of e401k, nettfa and marr on the odds of participation even if they are not statistically significant.

b. Explain how you reached Model 2.

c. Discuss which model is a better model in explaining the variability in the probability of participation.

d. Predict the probability of participation for the first 10 observations in the data set.

e. Compare the predictive accuracy of the two models you built in Q3 using 10-fold cross-validation. Discuss which model you would pick as a predictive model. Clearly state what measures you are using to pick your model and why?