The mixture of probability distribution functions by absolutely continuous weight functions

From MaRDI portal





Denote by E(a,b) the set of continuous probability distribution functions, which are strictly monotone increasing in [a,b] and have values 0 and 1 at the points a and b, resp. Let F(z)\(\in E(a,b)\) and the family of probability distribution functions G(z,x)\(\in E(a,b)\) with parameter \(x\in [0,1]\) be given. Assume that q(x) is an absolutely continuous probability function with probability density function f(x), \(x\in [0,1].\) Necessary and sufficient conditions are formulated (in two theorems) for the Fredholm's integral equation of first kind \[ \int^{1}_{0}G(z,x)f(x)dx=F(z) \] to have a solution with square integrable density function f(x). To solve this integral equation, it is traced back to the solution of the minimum problem of a symmetric definite Hilbert-Schmidt's kernel, using a distance concept between two probability distribution functions. The existence is shown of a family of probability functions, by which the given probability distribution function is not representable as their mixture.











This page was built for publication: The mixture of probability distribution functions by absolutely continuous weight functions

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3723409)