Greetings,
I have experimented with the MBESS and pwr packages for the estimation of
sample size for a given CV, precision, and confidence interval.
Thus far I have found the ss.aipe.cv {MBESS} (Sample size planning for the
coefficient of variation given the goal of Accuracy in Parameter Estimation
approach to sample size planning.) function to be best suited for my needs.
However, the data from which I am calculating my CV is approximately
log-normally distributed- and thus has a large CV (1.4). Using this CV,
precision (20% within the pop mean) and confidence interval (95%) parameters
I obviously get a suggested sample size that is very large (n = 1182). By
reducing my precision and confidence interval requirements to something like:
ss.aipe.cv(C.of.V=1.4, width=0.5, conf.level=0.9)
... the function still suggests about 230 samples which is near the upper
limit of feasibility.
I would like to deduce an optimal number of samples, however the log-normal
distribution of this data suggests that the above approach is not well suited
to this task.
Are there any better approaches or references which might send me in the right
direction?
Thanks in advance,
--
Dylan Beaudette
Soils and Biogeochemistry Graduate Group
University of California at Davis
530.754.7341