Preliminary analyses in support of the CEMP Review Workshop: power analyses

A series of power (sensitivity) analyses of CEMP indices was made using the DOS-based software MONITOR (J. P. Gibbs, 1995). In all, 170 time series were investigated for their power to detect change over the range –10% to +10% of a mean survey value. Each run of MONITOR involved the calculation of power for 3 survey scenarios and 3 alpha levels (0.05, 0.10 and 0.20). The 3 surveys scenarios were chosen to represent time series data from annual surveys conducted annually over periods of 5, 10 and 20 years.
Generally, time series arising from 5 annual surveys provided little power to detect change, and CEMP indices A3 (delta log transform), A8a (no transform), A8c (log odds transform), B1a (delta log transform), B1b (log odds transform) and C2b (no transform) performed poorly as indicator of change. However, changes within the range –10% to +10% can be detected in indices A1, A2, A5a, A7, A8b, C1 under the survey scenarios of 10 and 20 years and alpha levels of 0.05 or greater.