Discussion: View Thread

This message has been cross posted to the following eGroups: Statistics in Epidemiology Section and Statistical Consulting Section .
-------------------------------------------

I would appreciate input regarding the analytic strategy for this study.  Study design:  there are 3 types of clinics: A, B and C.  The population had very few type A and type B clinics, all of which were included in the sample by design (probability of selection=1).  Clinic type C was very common.  70% of type C clinics were selected using stratified random sampling by region (n=5 regions) to ensure that the sample proportion was similar to the population proportion of type C clinics within each region.   The analysis will be at the clinic level, and I anticipate a low (ignorable) correlation among clinics within each region.  My questions are: 1) Is it appropriate to create sample weights to account for the differential probability of selection?  2) If sample weights should be developed, I would be grateful for recommendations on how to create them.

Thank you!

-------------------------------------------
Amy Storfer-Isser, Ph.D.
Statistical Research Consultants, LLC
amysi.stats@gmail.com
-------------------------------------------

Hi Amy,

I think we need to know a bit more about your study, including the following:

- What differentiates clinic type C from clinic types A and B?  What differentiates clinic type B from clinic type A? 

- Originally, was there any interest in comparing study findings across clinic types?   What was the motivation for including clinics of type A and B in the study, even though there are very few such clinics?     

- What is the underlying purpose of the study?  Is it to compare a treatment group and a control group in terms of one or more outcomes collected at the clinic level? Or to understand more about how a clinic-level outcome varies over time and/or space? 

Thanks,

Isabella 

-------------------------------------------
Isabella Ghement
Ghement Statistical Consulting Company Ltd.
-------------------------------------------