Chapter 5: 5+, 6*


5(a) Show only your coefficient estimates, SEs, z- and p-values. 
     State an interpretation of the coefficients in context.
5(b-i)   [nothing to show] set.seed(23) just before using sample();  
         Use a 50-50 split for the training/validation set
5(b-ii)  [nothing to show] 
5(b-iii) [nothing to show] 
5(b-iv)  Show the confusion matrix and the computation of the validation set error.
5(c) Show only the validation set error for each split and comment on the results.
5(d)+set.seed(23) just before using sample() with a 50-50 split
     Show the confusion matrix and validation set error. In addition to the comment
     requested in the text, comment on whether this change (or lack of change) can
     be used to asses the significance of the variable 'student' in the model.

6(a) This is the same as #5(a), so you can skip it.
6(b) Show the code for your function
6(c) set.seed(1) just before calling boot(). Use 1,000 bootstrap replicates 
6(d) set.seed(1) just before calling boot(). Use 1,000 bootstrap replicates 
6(e)*Determine the median income and its bootstrap standard error.
     You will need to write a median.fn() function [use set.seed(1)]
     Show your code for this.
6(f)*Write code calling your median.fn() function 1,000 times to compute a 
     bootstrap estimate & SE "by hand" (in R) 
     set.seed(1) before the first call to sample(). 
     Report the bootstrap estimate and its standard error & show your code for this.
6(g)*Determine a 90% bootstrap confidence interval for the median as follows:
   i a percentile CI based on the $t replicate values (without using boot.ci)
  ii a Normal CI based on the output from boot() (without using boot.ci) 
 iii compare the above results to those from boot.ci()

**NOTE: boot.ci() can throw an error related to the BCa version. If this causes problems
when knitting an .Rmd file, you can use the argument type=c("norm","basic","perc"), or any 
subset, for the Normal, Basic, and Percentile versions avoiding the BCa version.