Tarone (1990, Biometrics 46, 515-522) proposed a multiple test procedu
re for discrete test statistics improving the usual Bonferroni procedu
re. However, Tarone's procedure is not monotone depending on the prede
termined multiple level alpha. Roth (1998, Journal of Statistical Plan
ning and inference, in press) developed a monotone version of Tarone's
procedure. We present a similar procedure that is both monotone and a
n improvement of Tarone's proposal. Based on this extension, we derive
a step-down procedure that is a corresponding improvement of Helm's (
1979, Scandinavian Journal of Statistics 6, 65-70) sequentially reject
ive procedure. It is shown how adjusted p-values can be computed for t
he monotone procedures. Finally, the described ideas are applied to th
e concept of Ruger tests.