Missing data estimation in tyrannosaurid dinosaurs: Can diameter take the place of circumference?

Missing data, due to taphonomic deformation, inaccessibility of specimens, or human error in collecting, cataloguing and measuring features, is a formidable problem in current palaeontological studies. Missing values within a data set can undermine confidence in analyses, skew results in promoting analyses of small portions of a population, not necessarily representative of the entire data set, and drastically decrease sample sizes. Missing data estimation methods, however, may reduce the effects of these missing values and potentially boost sample sizes for palaeontological studies. Here, six missing data estimation models for the prediction of femoral circumferences in tyrannosaurids were statistically tested and their predictive success measured against true circumferences, and other models. The statistical analyses suggest that estimation models based on anteroposterior diameter values of tyrannosaurid femora were poor predictors of circumference, whereas those based on mediolateral diameters were much more successful. Three out of the six models, were presented as viable alternatives to missing measured circumferences and may be used to boost tyrannosaurid samples with significant levels of missing data.