.It is well known that kernel density estimations are popular estimators of an unknown density function. But they suffer boundary bias for density functions that have finite or semi-finite supports especially nonnegative support. In this thesis, for such densities with nonnegative supports, we discuss three recent kernel estimations that use gamma and beta prime densities as kernels. These estimators no longer have boundary bias. Some theoretical properties of such estimators are established. Also, through Monte Carlo simulation, comparisons between introduced estimators and existing estimators are carried out
