Reviewing Green et al. (2010)

Green et al. (2010) used random assignment to assess a parent based intervention for children diagnosed with autism. The study is being discussed as a marked improvement over current research in the autism field. I disagree.

With this study you get:

-Random assignment

-sample (n) >30

And that is it for advantages. I honestly think people see the words "random assignment" and a big (n) size and put this study on a pedestal. It doesn’t deserve it, even compared to what is currently available.

In group design there are the following threats to internal validity:

-History (controlled in this study)
-Experimental mortality (controlled in this study)
-Regression toward the mean (controlled in this study)
-Maturation (a problem in this study)
-Instrumentation (a problem in this study)
-Testing Bias (a problem in this study)
-Instrumentation (a problem in this study)
-Selection Bias (controlled)

I don’t see how the authors can claim fidelity of the intervention. The parents were applying the intervention (or not) as the case might be. Occasional direct measures a la the videotapes or clinical visits are no guarantees of fidelity. This is a not a minor problem when doing consultation. It is a huge problem. Also, non-blinded terminal testing such as the ADOS is a concern. Where are the direct behavioral measures?

This is not a true experimental design; this is an indirect study that uses random assignment. The use of random assignment may be an anomaly, but it is not a step forward in this case.

Notes: For a different view of this research go see Michelle's analysis.