Summary
Special emphasis should be placed on determining optimal sampling regimes to detect statistically significant differences in key indicators between populations for species with high conservation values, especially when invasive handling techniques are the only method of gathering data. In procellariiform nestlings one such key indicator is peak mass, which serves as a measure of parental investment and is an important factor for future survival. Daily measurements of mass for 15 wandering albatross Diomedea exulans chicks ranging in age from 14 to 270 d were obtained in 1963 on Bird Island, South Georgia. Log-transformed mass was modelled using a linear mixed model (LMM) that included a linear term for age combined with smooth nonlinear departures estimated as random effects, random chick effects and a continuous-time autoregressive error process. The relative efficiency of different sample sizes of chicks versus frequency of measurement was investigated by fitting this LMM to sub-samples of the dataset of either 5, 10 or 15 chicks with the corresponding frequency manipulated to give the same total number of measurements for each sub-sample. Measures of precision of predictions from the LMM obtained for each dataset included (1) the width of single standard error bounds about predicted peak mass as a percentage of this mass (PSEPM), and (2) the probability of detecting a P-percentage difference in peak mass between 2 theoretical populations for the same sample size of chicks and same frequency of measurement in each (PRDPM). The most efficient of these sampling regimes for detecting a difference between populations or cohorts was 15 birds sampled every third day, which gave a PRDPM of 0.95 for a P of 15%.