×

Simulation studies on robustness of the t- and u-test against truncation of the normal distribution. (English) Zbl 0594.62037

Robustness of statistical methods and nonparametric statistics, Pap. Conf., Schwerin/Mecklenb. 1983, Theory Decis. Libr., Ser. B 1, 145-151 (1984).
[For the entire collection see Zbl 0588.00027.]
The aim of this simulation-study was to determine the influence of the truncation of a normal-distributed random variable (in practice occuring e.g. by selection) on the power functions of three tests, the t-test, the u-test and a modified Johnson-test. Particulary we were interested in finding sample sizes ensuring a good behaviour of the tests. More precisely, we looked for an \(n^*\) so that a test is ”\(\epsilon\)-robust” for \(n>n^*\).

MSC:

62F35 Robustness and adaptive procedures (parametric inference)
65C99 Probabilistic methods, stochastic differential equations

Citations:

Zbl 0588.00027