Customer Reviews

Design and Analysis of Group- Randomized Trials

5 star:		(0)
4 star:		(1)
3 star:		(0)
2 star:		(0)
1 star:		(0)

 

 

Group-randomized trials are trials in which randomization takes place at a level higher than the individual subject. For example, for logistical reasons a comparison of two programs might be undertaken by randomizing communities to program, in which case the community rather than the resident is the unit of analysis.

Even though Murray's text was published in 1998, it is still the outstanding book in the field. (Are there any others?) You will find reading this book to be a profitable experience if only because you can point to it (literally!) when telling someone that the number of observations is the number of CITIES that were randomized to be targeted with an anti-smoking campaign or not and NOT the number of RESIDENTS. The writing is clear and straightforward. To get the most out of it you should be comfortable with with mathematical models (use the Search Inside facility to look at the bottom of page 140, for example), which are used to describe things unambiguously rather than to develop mathemeatical theory. Even if you are not comfortable with mathematical models, there is a lot to be gained from Murray's discussion of the issues.

The flaw is MAJOR. The book contains numerous analyses of data from the Minnesota Hearth Health Program. However, Murray is not able to provide the original data. With the original data, one could be an active learner by replicationg his analyses and trying other aproaches to see what effects they might have. Without the data, learning is much more passive.

I struggled hard choosing the number of stars. I thought about 3 starts because, when you get right down to it, the analysis of group-randomized trials livolves nothing more than a particular kind of hierarchical model, which other texts discuss splendidly. However, if someone wanted to know about group randomized trials specifically, this is where I would send him/her, hence the 4 stars. If the data were available, the rating would be 5 stars without question.