PDF pending restoration
ERIC Number: ED144944
Record Type: RIE
Publication Date: 1975-Sep
Reference Count: 0
A Developmental Perspective on Statistical Methods of Removing Bias in Quasi-Experiments. Final Report.
Bryk, Anthony S.; Weisberg, Herbert I.
A mathematical model was developed to represent the way data are generated in quasi-experiments designed to study interventions in growth systems. This model explicitly represents three important processes: (1) the nature of individual growth; (2) the process by which individuals are assigned to treatment groups; and (3) the nature of the treatment effect. The way in which these three factors relate to the effectiveness of various methods is explored via both mathematical analysis and computer simulation. Results suggest that, in general, none of the statistical methods is very effective in removing the biases resulting from non-random assignment. It is concluded that the traditional pre-post, treatment versus control design with non-random assignment is not adequate in growth situations, and that research should focus on the development of experimental designs and corresponding analyses better suited to the dynamic nature of developmental processes. The FORTRAN IV computer program used for the simulation is appended. (Author/MV)
Publication Type: Reports - Research
Education Level: N/A
Sponsor: National Inst. of Education (DHEW), Washington, DC.
Authoring Institution: Contract Research Corp., Belmont, MA.
Note: Best copy available