In order to provide answers to the first question concerning minimum m,
simulations were run for three common values of ns, i.e., ns=5, ns=7 and ns=10.
delta_mu for these simulations was held at zero. That choice results in an in-control
simulation with the resulting probability of alarm representing the type I error rate α.
Runs were made using various values for. The probability of an alarm was recorded for
each method and each run. The methods simulated were that of Tsai, Nedumaran
&Pignatiellio and Albers &Kallenberg. For reference purposes, simulations were also
run with m=nq representing the distribution of probability of alarm for the standard
method with Quesenberry’s recommendation for m (labeled Q standard). For each set of
control values1000 replications were made.