CSI 779, Spring 1995, J. E. Gentle

Assignment due April 10

This assignment requires you to get some software over the net.
I just checked the turnaround time (~15 seconds), and the file sizes.
The total amount does not take up a lot of space, so the assignment
requires everyone to get it, even though it would be more efficient to
have one copy for all.  I suggest that you create a subdirectory, say
bootci, in which you build the software and do the assignment.
That way you'll have all the files together and you may want to delete
all of them after awhile.

To get the software, send email to statlib@lib.stat.cmu.edu with the
message
   send bootstrap.funs from S
(This is described in the Appendix of Efron and Tibshirani.)
The software comes in two shar files.  (I'm not sure why they were
broken up; you only need the first one, however.)  Look at the file in
an editor, and delete the lines at the beginning (to just after the 
line that says ...  CUT HERE ...).  Put the file in your bootstrap
directory, and unpack it by 
   sh file_name
(These instructions are in the packed file, as comments at the beginning.)

This will create a README file.  Read it, and follow the instructions in
it.  (You do not need to compile and link the Fortran functions, (i.e.,
skip Step 1).)  Test abcnon using the example on p. 401 of Efron and
Tibshirani.  The command help(bootstrap) may not work because the help
files want to use nroff -- but who needs help?

1. (Similar to 13.4)  Let (x_1, x_2, ..., x_20) be a random sample from
  an exponential distribution with mean theta=1. (Use rexp in Splus)
  Based on a Monte Carlo sample size of 400 and bootstrap sizes of 200,
  construct a table similar to Table 13.3 for coverages of 95% confidence
  intervals based on 
    standard normal
    nonparametric percentile
    BCa (use bcanon)
    nonparametric ABC (use abcnon)

2. Work Exercise 13.5, p. 177.

3. Work Exercise 14.6, p. 199.