An Empirical Investigation of Several Tests for the Mean of a First-Order Autoregressive Process.
DESMATICS INC STATE COLLEGE PA
Pagination or Media Count:
The study of the effects of unusual environments on individuals often entails the analysis of repeated measurements taken on a single subject. Unfortunately, except under very restrictive sets of assumptions, no valid statistical techniques have been developed for such an analysis. In particular, for the first-order autoregressive process no inference procedure is currently available which enables the analyst to control the probability of making an invalid conclusion. This problem is particularly acute when only a relatively small number of observations are available for the analysis. The purpose of this investigation is to evaluate some of the procedures which have been suggested for this situation. Four test statistics are considered for testing hypotheses about the mean of this autoregressive process. Of particular interest is the difference between the nominal error rate chosen by the experimenter and the actual error rate given by thr procedure. It is also desirable to evaluate how this difference is affected by the number of observations used in the analysis.
- Statistics and Probability