A STUDY OF THE EFFECT OF PROXIMALLY AUTOCORRELATELD ERROR ON TESTS OF SIGNIFICANCE FOR THE INTERRUPTED TIME SERIES QUASIEXPFILImENIAL DESIGN*

The primary purpose of the present study was toinvestigate the appropriateness of several tests of significance for use with interrupted time series data. The second purpose was todetermine what effect the violation of the assumption of uncorrelated error would have on the three tests of significance. The three tests were the Mood test, Walker-Lev Test 3, and Double Extrapolation Technique. The procedure was basically that of generating a large number of time series having specified characteristics and performing the tests of significance on each generated time series. The results of the study indicated that the three tests of significance are appropriate for use on data of interrupted time series form. Tables and figures illustrate the text. (Author/DB) U.S. DEPARTMENT OF HEALTH