Re: radnomizing a sub-sample

[<DavidR46@xxxxxxx>]

Here are some questions to consider.

 

Do you want to generalize to all cases in Toronto, or some other geographic area?

 

Do you think that there may be features of assigned social workers that could account for some findings.  Is it likely that worker characteristics may have an effect on outcome variables of interest?

 

Have you considered a cluster randomize sample, in which clients are "clustered" within workers, workers are clustered within offices, offices are clustered within area authorities?

 

What effect size do you anticipate?

 

What is the power of your statistical test? 

 

Sometimes the answers to these questions may influence your sampling planning.

 

David

 

 

David S. Robinson, Ed.D.
President
DSRobinson & Associates
903 Providence Place, #155
Voice: 401-383-7953
Mobile: 617-733-5979
Email: drobinson@xxxxxxxxxxxxxxxxxx
Web: www.evaluationhelp.com

---

Get a sneak peek of the all-new AOL.com.