Key points are not available for this paper at this time.
The authors argue that large scale educational interventions at the national level should incorporate evaluation as an integral part of program development. They suggest several statistical techniques which developers could use. Until now, the thrust of evaluation has been to take a total program and judge it as either worthy or unworthy. In contrast, the authors propose an approach which searches for those few versions which are working well, so that they can be improved and recreated in future centers. The Head Start experience illustrates many of the authors' points.
Light et al. (Wed,) studied this question.