Search

A Monte Carlo study was done to determine the adequate sample size for quasi-experimental regression studies, which compare regression lines for two groups and estimate their point of intersection. Populations of 1,000 subjects in each of two groups were constructed (using random normal deviates) to yield equivalent regression lines of opposite slope. Multiple runs of subset samples of 20, 25, 30, 40, and 50 points per group provided the sampling error of estimating X- and Y-coordinates for the intersection point. Standard error was an inverse function of group size (degrees of freedom), as with other sampling statistics. The function leveled off at about 50 subjects per group, suggesting this as the recommended minimal group size for such analyses. The author notes that the acceptable sample size for estimating the intersection of two lines (N=50/group) exceeds that for estimating the difference between sample means (N=30/group) because of the additional parameter involved. (LPG)