The RMC Model A (norm-referenced) for evaluation of Title I programs is based upon the equipercentile assumption--that students maintain their percentile rank over a one-year period, provided that no special instrucional intervention is introduced. The control group, essentially the sample used to standardize the achievement test, represents the no-treatment expectation. The treatment effect is defined as the average posttest performance minus the average pretest performance. Title I participants are usually selected based upon low pretest scores, and may bias the results of Model A evaluation because low ability students generally regress toward the mean without any intervention. Two models which could estimate the no-treatment expectation and which might solve these problems were investigated: (1) Murray's recursive path analytic paradigm, encompassing three achievement measures, a group factor, an individual differences within group factor, and an error component; and (2) an extension of the one independent variable regression situation. Results indicated that for local districts with fewer than 500 students per grade, the no-treatment expectation would probably be overestimated. These two methods would be acceptable for use in districts with 2,000 or more students per grade. The use of local correlations, based on non-Title I participants, was also suggested. (GDC)