ABSTRACT
Throughout this book we have emphasized on the benefits of minimum distance estimation based on disparities. At the same time, we have pointed out that in case of continuous models one is forced to use some form of nonparametric smoothing such as kernel density estimation to produce a continuous estimate of the true density. As a result, the minimum distance estimation method based on disparities inherits all the associated complications such as those related to bandwidth selection in continuous models. Basu and Lindsay (1994) considered another modification of this approach, discussed in Chapter 3, where the model is smoothed with the same kernel as the data to reduce the dependence of the procedure on the smoothing method. Although the Basu and Lindsay procedure largely reduces the effect of the bandwidth, the process still involves a nonparametric smoothing component.
