Abstract
In the absence of randomized experimentation, prevention intervention effects are difficult to isolate from sources of bias such as self-selection, differential attrition, etc. It is proposed that modern methods of structural equation modeling, such as can be done with the EQS computer program, can help to isolate true effects from observed effects, differentiate pretest differences from treatment effects, control for missing data, and evaluate effectiveness of treatment using indicators of the degree of program participation.