Estimating Gompertz Growth Curves from Marine Mammal Strandings in the Presence of Missing Data

Abstract

Stranded bottlenose dolphins (Tursiops truncatus) off the coast of South Carolina (SC) provide data essential for population health assessment. Of the 598 bottlenose dolphin strandings in SC from 1993 to 2007, 91 were of sufficient body condition to obtain organ weights. Of these 91 animals, only 52 were brought back to the laboratory for total body weight measurements. Because it is more feasible to transport smaller animals to the laboratory setting for necropsy procedures, a selection bias is present in that data for larger animals are often missing. Regression and propensity score multiple imputation methods are utilized to account for missing data needed to compute growth. Fitted Gompertz growth curves for SC animals with and without adjustment for missing data are compared to those found from the northwestern Gulf of Mexico. South Carolina animals display a trend in lower asymptotic mean total body weights and faster growth rates compared to the Gulf of Mexico population. The differences generally increased in magnitude after imputation methods. South Carolina females were originally estimated to reach larger maximum sizes than Gulf of Mexico females, but after imputation this relationship reversed. The findings suggest selection bias should be accounted for in sampling stranded dolphins.