title	author	date	output
Non-parametric estimation of mixing distribution and the proportion of zeros	Xiangjie Xue	06 三月, 2022	html_document keep_md true

This is a R package for estimating the mixing distribution and the proportion of null effects accompanying the doctoral thesis by Xiangjie Xue. The main webpage of this package is here.

Installation

The package can be installed via devtools or remotes:

devtools::install_github("xiangjiexue/npfixedcomp2")

NOTE: This package contains Fortran and C++ codes with C++11 standard, so make sure gfortran and gcc/clang is available.

Available family functions

Since this package is written in C++, there are limited number of family functions available.

Family	loss	original method	large-scale implementation
normal	maximum likelihood	npnormll	npnormllw
one-parameter normal	maximum likelihood	npnormcll	N/A (requires variational bin)
(non-central) t	maximum likelihood	nptll	nptllw
poisson	maximum likelihood	nppois	N/A (binning by default)
normal	Cramer-von Mises	npnormcvm	npnormcvmw
normal	Anderson-Darling	npnormad	npnormadw
bivariate normal (experimental)	maximum likelihood	npnorm2Dll	N/A (not yet implemented)

It is noted that after version 1.1, all the header files are implemented as template. It is possible to write implementation using other scalar type rather than double (float, long double, etc.) provided that it is supported by Eigen, but please make sure that the precision is what you want.

template<class Type>
class npfixedcomp
{
public:
  ...
};

template<class Type>
class npnormll : public npfixedcomp<Type>
{
public:
  ...
};