Criteria of goodness of fit test in the selection of non-linear models for the description of biological performances

Verena Torres, I. Barbosa, R. Meyer

Abstract


A study was undertaken to analyze the goodness of fit of non-linear models, for describing the live weight performance during growth of
crossbred Bufalypsos animals under grazing conditions through the utilization of fourteen statistical criteria. The non-linear models in
the logistic, Gompertz, von Bertalanffy and Brody parameters were adjusted to the data corresponding to 43 weighings carried out during
two lactations (1064). The best criteria to select non-linear models are: coefficient of determination R2, mean square of the prediction error
(MSPE), standard error of estimation, standard error of each parameter, analysis of residuals, mean absolute error (MAE) and percentage
of mean absolute error (PMAE). Estimations of parameters, as the analysis of residuals, are inefficient when working with the average of
observations. In particular, for buffalo females the logistic was the best fitting model.
Key words: non-linear models, goodness of fit test, statistical criteria

Full Text:

PDF

Refbacks

  • There are currently no refbacks.