APPLICATION OF THE MULTIPLE IMPUTATIONS METHOD IN THE MORPHOMETRIC STUDY OF THEROPOD TEETH

JORGE GUSTAVO MESO; JUAN IGNACIO CANALE; MATTIA ANTONIO BAIANO; LEONARDO SALGADO; FERNANDO ARCHUBY

Corrientes

Congreso; XXXII Jornadas Argentinas de Paleontología de Vertebrados; 2018

From the pioneering work of Currie and collaborators at the beginning of the 90?s, a large number of qualitative and quantitative parameters began to be established and systematized, with the aim of achieving a more precise classification of isolated theropod teeth. Due to exposure of these materials to taphonomic processes, the missing data are relatively common in the study of these elements. Such loss of information can lead to a drastic reduction of the original dataset, and therefore to biased conclusions. Currently, few statistical programs are available that allow the direct use of incomplete datasets to perform different multivariate methods, but all of them show a low statistical power. The objective of this work is to evaluate the quality of a database obtained after the application of the multiple imputation method to recover lost data. The most common approaches to deal with missing data have different consequences on the dataset and hence on the results of the analysis. ?Casewise/variable deletion? consists on removing the specimens or variables that have missing values. In ?Mean value imputation? missing values are replaced by their column average, resulting in a reduction of variation and an artificial increase in the power of tests. In the method of ?Iterative imputation?, missing values are initially replaced by their column average. An initial PCA run is then used to compute regression values for the missing data. The procedure isiterated until convergence. The results obtained suggest this method provides a better resolution for the identification of isolated teeth.