On page 15, column 1, the value associated with the 80th
precentile of the second customer's sojourn time for
an M/M/1 queue beginning empty and idle is reported as
1.935. Marvin Nakayama determined that this number is
a typographical error and should be 2.124, which he
verified by implementing the algorithms in the paper in
Matlab and also by Monte Carlo simulation. Matt Robinson
used APPLPy (see https://github.com/MthwRobinson/APPLPy/wiki
and https://pypi.python.org/pypi/APPLPy/) with the following
code and output.
for k in [0,3,6]:
X = ExponentialRV(Rational(1,1))
Y = ExponentialRV(Rational(10,9))
Z = Queue(X,Y,2,k,1)
pct = Z.variate(s=.8)
print pct[0], k
2.12376635506612 0
4.43171682160014 3
7.48902024554419 6