COMPARISON OF A PERFORMANCE-TEST BATTERY IMPLEMENTED ON DIFFERENT HARDWARE AND SOFTWARE - APTS VERSUS DELTA

Although every effort is made to make performance tests comparable, im plementing the same test or test battery on a different computer in a different program language can produce only approximately identical te sts because of differences in hardware and software. Therefore, compar ability of the tests is an empirical as opposed to theoretical issue. The present paper examined a battery of six performance tests implemen ted on two computers, a NEC laptop programmed in N82-BASIC, and a COMP AQ laptop programmed in TurboC. The asymptotic performance levels diff ered somewhat, as did the rate of approach to asymptote in some cases. More important, after correction for reliability attenuation, the cro sscorrelations between the two different implementations of the same t est approached unity, which supports the conclusion that what was meas ured after the tests had stabilized was identical for the two differen t hardware and software configurations. Furthermore, comparison of int ercorrelations among tests on the different computers showed no signif icant differences, which suggests that the factorial structure is the same.