glm.hpp

Go to the documentation of this file.

#ifndef GENESIS_UTILS_MATH_REGRESSION_GLM_H_
#define GENESIS_UTILS_MATH_REGRESSION_GLM_H_
 
/*
    Genesis - A toolkit for working with phylogenetic data.
    Copyright (C) 2014-2024 Lucas Czech
 
    This program is free software: you can redistribute it and/or modify
    it under the terms of the GNU General Public License as published by
    the Free Software Foundation, either version 3 of the License, or
    (at your option) any later version.
 
    This program is distributed in the hope that it will be useful,
    but WITHOUT ANY WARRANTY; without even the implied warranty of
    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
    GNU General Public License for more details.
 
    You should have received a copy of the GNU General Public License
    along with this program.  If not, see <http://www.gnu.org/licenses/>.
 
    Contact:
    Lucas Czech <lucas.czech@h-its.org>
    Exelixis Lab, Heidelberg Institute for Theoretical Studies
    Schloss-Wolfsbrunnenweg 35, D-69118 Heidelberg, Germany
*/
 
#include "genesis/utils/containers/matrix.hpp"
#include "genesis/utils/math/regression/family.hpp"
#include "genesis/utils/math/regression/link.hpp"
 
#include <utility>
#include <vector>
 
namespace genesis {
namespace utils {
 
// =================================================================================================
//     GLM Data Structures
// =================================================================================================
 
struct GlmExtras
{
    std::vector<double> initial_fittings;
    std::vector<double> prior_weights;
    bool                with_intercept = true;
 
    std::vector<size_t> strata;
 
    enum ResidualType
    {
        kDefault,
        kPearsonResiduals,
        kDevianceResiduals
    };
 
    ResidualType residual_type = ResidualType::kDefault;
 
    bool mean_deviance = false;
};
 
struct GlmControl
{
    size_t max_iterations = 25;
 
    double epsilon = 1.e-5;
 
    double max_r2 = 0.99;
};
 
struct GlmOutput
{
    bool converged = false;
    size_t num_iterations = 0;
 
    size_t rank = 0;
 
    size_t df_resid = 0;
 
    double scale = 1.0;
 
    Matrix<double> Xb;
 
    std::vector<double> fitted;
 
    std::vector<double> resid;
 
    std::vector<double> weights;
 
    std::vector<double> which;
 
    std::vector<double> betaQ;
 
    std::vector<double> tri;
 
    double null_deviance = 0.0;
 
    double deviance = 0.0;
};
 
// =================================================================================================
//     GLM Fit
// =================================================================================================
 
GlmOutput glm_fit(
    Matrix<double> const&      x_predictors,
    std::vector<double> const& y_response,
    GlmFamily const&           family,
    GlmLink const&             link,
    GlmExtras const&           extras = {},
    GlmControl const&          control = {}
);
 
GlmOutput glm_fit(
    Matrix<double> const&      x_predictors,
    std::vector<double> const& y_response,
    GlmFamily const&           family,
    GlmExtras const&           extras = {},
    GlmControl const&          control = {}
);
 
GlmOutput glm_fit(
    Matrix<double> const&      x_predictors,
    std::vector<double> const& y_response,
    GlmExtras const&           extras = {},
    GlmControl const&          control = {}
);
 
// =================================================================================================
//     GLM Output
// =================================================================================================
 
std::vector<double> glm_estimate_betas( GlmOutput const& output );
 
// /**
//  * @brief Obtain beta estimates and variance covariance matrix of estimates from output the output
//  * of glm_fit().
//  *
//  * The resulting variance covariance matrix is a packed symmetric matrix with the size of the
//  * number of predictor variables (which is the size of the betas).
//  * Robust variance is calculated if the "meat" matrix for the information sandwich is supplied.
//  */
// std::pair<std::vector<double>, std::vector<double>> glm_estimate_betas_and_var_covar(
//     GlmOutput const& output,
//     std::vector<double> const& meat = std::vector<double>{}
// );
 
double glm_estimate_intercept(
    Matrix<double> const&      x_predictors,
    std::vector<double> const& y_response,
    GlmOutput const&           output,
    std::vector<double> const& betas
);
 
double glm_estimate_intercept(
    Matrix<double> const&      x_predictors,
    std::vector<double> const& y_response,
    GlmLink const&             link,
    GlmOutput const&           output,
    std::vector<double> const& betas
);
 
std::vector<double> glm_coefficients(
    Matrix<double> const&      x_predictors,
    std::vector<double> const& y_response,
    GlmOutput const&           output
);
 
std::vector<double> glm_coefficients(
    Matrix<double> const&      x_predictors,
    std::vector<double> const& y_response,
    GlmLink const&             link,
    GlmOutput const&           output
);
 
} // namespace utils
} // namespace genesis
 
#endif // include guard