Uma modificação da extensão do algoritmo AID para modelos lineares generalizados usando reamostragem Bootstrap
Presotti, Cátia Valéria
MetadataShow full item record
One of the most frequently situation found by researchers is to find groups of similar individuals. The cluster analysis is a set of statistical techniques that identify mutually exclusive subgroups or classes over individuals, based on their similarity. When then main is to group means of treatments we can use contrasts, multiple comparitions or clustering techniques as the SCOTT-KNOTT test and AID (Automatic Interaction Detector) technique. In this work we focus on the comparition of the simulated power function of the asymptotic test and also of the bootstrap test for the extension of the AID algorithm for generalized linear models. The bootstrap power function over main the asymptotic power when the number of binomial sample is equal to one, and the number of treatments is equal to 8 and 12, in a completed randomized experiment with a single factor for binomial variables.