When outliers infect data, the quasi-likelihood estimation will be affected and the fitted results will be inadequate. In this paper, we propose a fuzzy-weighted estimation that can reduce the influence of outliers efficiently. These fuzzy weights are computed by using the optimal fuzzy clustering method. Moreover, we call this new estimation procedure as fuzzy-weighted quasi-likelihood estimation. Under some weak conditions, we show the fuzzy-weighted estimators are √n-consistent and asymptotically normal. Moreover, a practical example and some Monte Carlo simulations are used to demonstrate the application of the fuzzy-weighted estimation method.
Relation:
International Journal of Information and Management Sciences 11(1), pp.65-73