Date of Award

8-1981

Degree Name

Doctor of Education

Department

Educational Leadership

First Advisor

Dr. Mary Anne Bunda

Second Advisor

Dr. Bradley Huitema

Third Advisor

Dr. James Sanders

Abstract

The focus of this study was the tenability of the underlying assumption of Model A, the norm-referenced model, of the Title I Evaluation and Reporting System (TIERS). Model A estimates the no-treatment expectation from normative data. The assumption made by the Model is that, in the absence of special treatment, student groups maintain a constant percentile position in relation to national norms.

The study examined the extent to which student groups maintain a constant percentile position over a six month period. The study used a sizable representative student subsample of the standardization group for the 1978 Metropolitan Achievement Test (MAT). The study design included replications of data analyses for four different grade levels on two MAT subject area test scores. Class means were used as the unit of analysis. The study sample contained approximately 250 classes per grade level.

Constancy of class percentile rank was examined for grades 2, 3, 4, and 5 on MAT reading and math tests using the normal curve equivalent (NCE) score metric. Analyses were conducted using NCE scores calculated according to TIERS guidelines. A second supplemental series of analyses were conducted for NCE scores derived from the percentile distributions of MAT class mean scale scores of the sample.

Study results indicated classes generally do not maintain a constant percentile position in relation to national norms. Class means were found to vary considerably from fall to spring in relation to national norms. The degree to which class means maintained their position fall to spring was more variable in math than in reading and changes in class mean position fall to spring were slightly larger in math than in reading. These two patterns held in each of four grade levels examined in this study. No distinct between-grade-level patterns were apparent.

For the four grades combined, the 95% confidence interval for class change scores in reading was estimated at -6.3 to +9.3 NCEs; the 95% confidence interval for class change scores in math was estimated at -9.8 to +13.8 NCEs. Note that zero growth is not at the center of the interval. The degree to which classes maintained their position in relation to the group score distributions of the sample in the supplemental analysis was even more variable than the degree to which classes maintained their position in relation to the MAT test norms for individuals.

The no-treatment expectation under Model A is based on the assumption that student groups maintain a constant percentile position. Based on the results of this study, there is no reason to assume student groups maintain a constant percentile position as a no-treatment expectation. Users of the Model, given the results of this study, should not view Model A change scores as a meaningful indicator of the quality of their educational program.

Access Setting

Dissertation-Open Access

Share

COinS