mig_rlcv function

Robust likelihood cross-validation for kernel density estimation for MIG

Given a data matrix over a half-space defined by beta, compute the log density using leave-one-out cross validation, taking in turn an observation as location vector and computing the density of the resulting mixture.

mig_rlcv(x, beta, Omega, an, xsamp, dxsamp, mckern = TRUE)

Arguments

• x: n by d matrix of quantiles
• beta: d vector $\boldsymbol{\beta}$ defining the half-space through $\boldsymbol{\beta}^{\top}\boldsymbol{\xi}>0$
• Omega: d by d positive definite scale matrix $\boldsymbol{\Omega}$
• xsamp: matrix of points at which to evaluate the integral
• dxsamp: density of points

Returns

the value of the likelihood cross-validation criterion