|
Abstract:
|
Recent public health interest has focused on the use of a 99th percentile for Body Mass Index as a measure of childhood obesity. Flegal, et al has shown an extrapolation of the CDC growth curves to estimate the 99th percentile is inconsistent with an empirical estimation of the 99th percentile. Waterlow et al and Guo et al have examined minimum sample size criteria for growth data, but they did not consider extreme percentiles. In clinical chemistry publications, there is a range in the minimum sample size related to non-normality of the data and to the (arbitrary) criteria used to determine minimum sample size. In this presentation, national data are used as the basis of simulations to investigate a minimum sample size criteria for the computation of a 99th percentile for BMI. Further, the sensitivity of the recommended sample size related to skewness and kurtosis is investigated.
|
