Regression.cpp File Reference

#include "Stats/Regression.hpp"
#include "Db/Db.hpp"
#include "Model/Model.hpp"

Functions
bool	_regressionCheck (Db *db1, int icol0, const VectorInt &icols, int mode, Db *db2, const Model *model)
bool	_regressionLoad (Db *db1, Db *db2, int iech, int icol0, const VectorInt &icols, int mode, int flagCst, const Model *model, double *value, VectorDouble &x)
VectorDouble	regressionDeming (const VectorDouble &x, const VectorDouble &y, double delta)
Regression	regression (Db *db1, const String &nameResp, const VectorString &nameAux, int mode, bool flagCst, Db *db2, const Model *model)

Function Documentation

bool _regressionCheck	(	Db *	db1,
		int	icol0,
		const VectorInt &	icols,
		int	mode,
		Db *	db2,
		const Model *	model
	)

bool _regressionLoad	(	Db *	db1,
		Db *	db2,
		int	iech,
		int	icol0,
		const VectorInt &	icols,
		int	mode,
		int	flagCst,
		const Model *	model,
		double *	value,
		VectorDouble &	x
	)

Regression regression	(	Db *	db1,
		const String &	nameResp,
		const VectorString &	nameAux,
		int	mode,
		bool	flagCst,
		Db *	db2,
		const Model *	model
	)

Evaluate the regression

Returns

Error return code

Parameters

[in,out]	db1	Db descriptor (for target variable)
[in]	nameResp	Name of the target variable
[in]	nameAux	Vector of names of the explanatory variables
[in]	mode	Type of calculation 0 : standard multivariate case 1 : using external drifts 2 : using standard drift functions (in 'model')
[in]	flagCst	The constant is added as explanatory variable
[in]	db2	Db descriptor (for auxiliary variables)
[in]	model	Model (only used for Drift functions if mode==2)

Remarks

The flag_mode indicates the type of regression calculation:

0 : V[icol] as a function of V[icols[i]]

1 : Z1 as a function of the different Fi's

VectorDouble regressionDeming	(	const VectorDouble &	x,
		const VectorDouble &	y,
		double	delta
	)

Calculate the coefficients of the Deming regression (with 2 variables)

Parameters

x	Vector for the first variable
y	Vector for the second variable
delta	ratio of error variances (s_y^2 / s_x^2)

Returns

Vector of coefficients for the equation

y = beta[0] + beta[1] * x

Remarks

Both input vectors are assumed to contain valid values

From: https://en.wikipedia.org/wiki/Deming_regression