Body mass estimation from cheek tooth measurements in extinct caviomorphs (Ctenohystrica, Hystricognathi): the importance of predictor, reference sample and method

Caviomorph rodents have presented an astonishing size variation both in the past and present times. The objective of this work is to estimate the body mass of 32 extinct caviomorphs and to analyze how several factors influence estimates: (1) the predictors selected (i.e., individual teeth and toothrows); (2) the reference sample used (i.e., rodents, rodents + artiodactyls, or rodents + artiodactyls + perissodactyls + hyracoids + lagomorphs); and (3) the method employed (i.e., simple vs. multiple regressions and whether taxonomic abundance and phylogeny are considered or not). The results indicate that the best dental structures to predict body mass are usually the upper and lower toothrow, second upper and lower molar, and first upper molar. Including ungulates improves predictions for living large rodents but not for all living caviomorphs or for small rodents. Thus, we propose different sets of best models depending on the size range of the extinct caviomorph under consideration. Ordinary least squares (OLS) regression, especially taxonomically weighted OLS, performs better in most cases. Multiple regression, associating dental length and width, usually improves the mean of the prediction error with respect to simple regression models. Despite a probable redundancy between these variables, independent information would be sufficient to improve the prediction error. Our estimates from dental dimensions for large and giant chinchilloids are within the range of values previously reported when a cranial variable that is considered as a robust body size proxy is used. For small extinct caviomorphs, our estimates deviate more from published body mass values.