Test for parameters equality in nonlinear regression models. I. Data in the randomized complete design

Adair José REGAZZI[1]

Carlos Henrique Osório SILVA1

§       ABSTRACT: We considerer the adjustment of g nonlinear regression equations and the likelihood ratio test with c2 approximation in order to test model identity, and we also compare the approximations given by F and c2 test statistics in a simulation study. For sufficiently large data sets (N ³ 120) we found the two approximations to perform roughly the same. For smaller sample sizes the F statistic approximation should be preferred since it yielded smaller type I error rates for N values investigated.

§       KEYWORDS: Growth curve; likelihood; hypothesis test.


[1] Departamento de Informática, Universidade Federal de Viçosa - UFV, CEP 36570-000, Viçosa, MG, Brasil. E-mail: adairreg@ ufv.br / chos@dpi.ufv.br.