IMPACTO DA VIOLAÇÃO DOS PRESSUPOSTOS DA ESTATÍSTICA PARAMÉTRICA EM ESTUDOS MORFOMÉTRICOS

Impact of violating parametric statistical assumptions in morphometric studies

Abstract

The aim of the study was to highlight the importance of characterizing data in terms of its probability distribution and the impact on the results using parametric tests to compare means when the data distribution model is mis-specified, based on studies of crustacean morphometry. The dataset consisted of abdominal length measurements of lobsters categorized by the following qualitative variables: sex and month. The data were analyzed in five stages: (1) comparison of means without characterizing the continuous random variable; (2) characterization of the continuous random variable in terms of its probability distribution; (3) data transformation and re-characterization of the continuous random variable; (4) application of the generalized linear model for comparison of factors; and (5) evaluation of the impact of violating parametric statistical assumptions on the comparison of the factors. The analysis of the results showed that the impact of assumption violations led to misleading conclusions, particularly with the use of the Student's t test, which is widely used in biologic organism research. Therefore, the importance of correctly applying available statistical tools ensure the reliability of the results and proper inference.

Keywords: Student’s t-test, Normality, Variance, Generalized linear model.