Abstract
Absence of normality and groups size inequality. Does it affect to autocorrelation estimation? We have undertaken a Monte Carlo simulation study to examine the behaviour of four procedures for the diagnosis of the first order autocorrelation: ρHCH , ρ J, ρ PP y ρWHS. We have used two correlation matrices structures, AR and ARH, in a Groups x Occasions (3xq) design unbalanced with normal distribution and balanced with six non-normal distributions. The results indicate that the groups size inequality affects only to ρPP procedure when the matrix is AR+ and to ρWHS procedure when the matrix is ARH. On the other hand, every procedures are sensible to the absence of normality, ρ HCH and ρ J are conservatives and ρWHS is liberal por AR+ matrices and every methods are conservative for AR- matrices, overcoat as much as smaller are q and nj and elder is ρ.
