SPURIOUS ASOCIATION IN LOGISTIC TIME SERIES BINARY REGRESSION MODELS

Spurious association in the linear regression model occurs when the independent variable makes an important contribution to explaining the variability of the response variable according to the hypothesis test of the parameter of the independent variable, even though the two variables have no relationship. In models with categorical response variable, the presence of spurious association is not reported in the literature. Therefore, in this study the existence of the phenomenon is shown empirically in logistic regression when the data are generated by different processes of time series that involve non-stationary series. The analysis of the results indicates that this phenomenon occurs when the generating mechanism of the response variable and of the explicative variable is non-stationary.