Abstract
The effects of magnification on sample statistics calculated by maximum likelihood, class analysis, and probit analysis methods have been studied for log-normal distributions having standard deviations between 0.5 and 1.1 (geometric standard deviations between 1.65 and 3.00). A lower limit of observation has been assumed so that sample truncation occurs. Theoretical results for infinitely large samples and experimental results for samples of size 100, 200, 500, and 1000 covering ten levels of magnification are presented and compared. The results include estimates of the standard errors on the means and variances as calculated by the different methods.