 1. To generate the many datasets for the g-test simulation I ran the
    following command.  Be warned, it took a couple of days to run:

      sh g-test-commands
      gnuplot best-errors.dat > best-errors.png
      gnuplot best-complete.dat > best-complete.png
      gnuplot low-conversion-errors.dat > low-conversion-errors.png
      gnuplot low-conversion-complete.dat > low-conversion-complete.png
      gnuplot low-lift-errors.dat > low-lift-errors.png
      gnuplot low-lift-complete.dat > low-lift-complete.png

 2. To generate the datasets for running multiple versions I ran the
    following commands:

      ./multiple-version-simulation 2 1000000 | sort | perl -ne '
          printf "%.6f\t%.7f\t%.5f\n", $./1000000, $_, $_/$.*1000000
            if $. < 20000 and $. > 1000
        ' > multiple-version-2.txt

      ./multiple-version-simulation 3 1000000 | sort | perl -ne '
          printf "%.6f\t%.7f\t%.5f\n", $./1000000, $_, $_/$.*1000000
            if $. < 20000 and $. > 1000
        ' > multiple-version-3.txt

      ./multiple-version-simulation 4 1000000 | sort | perl -ne '
          printf "%.6f\t%.7f\t%.5f\n", $./1000000, $_, $_/$.*1000000
            if $. < 20000 and $. > 1000
        ' > multiple-version-4.txt

      ./multiple-version-simulation 5 1000000 | sort | perl -ne '
          printf "%.6f\t%.7f\t%.5f\n", $./1000000, $_, $_/$.*1000000
            if $. < 20000 and $. > 1000
        ' > multiple-version-5.txt

      gnuplot multiple-versions.dat > multiple-versions.png

 3. To generate the graphs of the various functions:

      gnuplot standard-normal.dat > standard-normal.png
      gnuplot fair-coin.dat > fair-coin.png
      gnuplot fair-coin2.dat > fair-coin2.png
      gnuplot unfair-coin.dat > unfair-coin.png
      gnuplot unfair-coin2.dat > unfair-coin2.png

 4. Copy all png files into img directory.
