Bootstrap enhanced model specification testing

This paper proposes a bootstrap method to enhance the performance of theinformation matrix test, and more in general of the score test. The informationmatrix test is a model specification test proposed by White (1982). The standardbootstrap method is to use the bootstrap distribution of the test statistic to obtaina critical value which is more accurate than the asymptotic critical value.However, the score test uses a quadratic form statistic. In the construction andimplementation of such a quadratic form statistic two important aspects whichdetermine the performance of the test (both under the null and the alternative),are (i) the weighting matrix (the estimate of the variance matrix) and (ii) thecritical value. In this paper the bootstrap is used to get simultaneously a bettervariance matrix estimate and accurate critical values. The information matrix testis studied in some Monte Carlo experiments. see above see above