TY - JOUR
AU - DETMANN, Edenio
AU - BONFÁ, Hugo Colombarolli
AU - CECON, Paulo Roberto
AU - SILVA, Fabyano Fonseca e
PY - 2017/12/22
Y2 - 2024/05/30
TI - ASSESSMENT OF HYPOTHESES TEST FOR THE COMPONENTS OF MEAN SQUARE ERROR OF PREDICTION
JF - Brazilian Journal of Biometrics
JA - Braz. J. Biom.
VL - 35
IS - 4
SE - Articles
DO -
UR - https://biometria.ufla.br/index.php/BBJ/article/view/135
SP - 658-669
AB - <p>The analysis of mean square error of prediction is helpful to compare measured values with values simulated by mathematical models. Such analysis is based on the orthogonal decomposition of this quantity into three components which will indicate the probable constraints of the model concerning bias, unequal variance, and incomplete covariation when contrasted to actual values. However, such analysis has been carried out as a descriptive procedure without an adequate hypotheses test framework. Thus, we aimed to develop single hypothesis test to evaluate the statistical significance of mean square error of prediction components based on likelihood ratio test and χ² distribution. This proposal was evaluated by using simulated populations and was applied to a dataset obtained by simulating characteristics of cattle diets using two different models. We concluded that this test might help the modeler to focus on the real significant constraints of his model and to work on doing the necessary modifications on its mathematical structure in order to improve the accuracy and precision of the simulated values.</p>
ER -