Patrick Sequeria writes:
>Hello from snowy Switzerland.
>
>At the initial Protocol Writing Workshop mention
>was made of the problems of randomising dental studies.
>
>In short what should the unit of randomisation be ?
>
>Participants / Teeth / Roots / Canals
It is snowy here in Missouri also, but not nearly as scenic.
The logistics of the study will determine at what level you randomize.
Generally, randomizing at the highest level (here the patient) simplifies
the analysis and often the logistics. Randomizing at a lower level is more
complex from both a statistical and a logistical perspective, but can often
lead to a substantial improvement in power.
If every patient has exactly two root canals, you could conceive of this as
a crossover design which is clearly more efficient. It would also have some
of the same analysis and interpretation issues that a crossover design would
have.
The statistical models that allow for correlation between multiple teeth in
the same mouth are rather tricky, so you definitely want to consult a
statistician early in the process. I'd love to help if you can fly me out to
Switzerland, but you probably can find someone closer. This is the sort of
thing where a face-to-face consultation is far superior to consulting via
email.
Good luck!
Steve Simon, [log in to unmask], Standard Disclaimer.
STATS: STeve's Attempt to Teach Statistics. http://www.cmh.edu/stats
|