`regress` (Series)

Ordinary or weighted least-square regression

Syntax

[res, stdEst, res, stdRes, fit, dates, covEst] = regress(lhs, rhs, ...)

Input arguments

lhs [ Series ]

Time series of dependent (LHS) variables; can be a multivariate time series object for multiple dependent variables (sharing the same explanatory variables).

rhs [ Series ]

Time series of explanatory (RHS) variables; can be a multivariate time series object for multiple explanatory variables.

Output arguments

est [ numeric ]

Vector of estimated regression parameters.

stdEst [ numeric ]

Vector of std errors of the parameter estimates.

res [ Series ]

Time series of the regression residuals.

stdRes [ numeric ]

Estimate of the std deviation of the regression residuals.

fit [ Series ]

Time series of fitted values for the LHS variable(s).

Dates [ numeric ]

The dates of observations actually used in the regression.

covEst [ numeric ]

Covariance matrix of the regression parameter estimates.

Options

Dates=Inf [ Dater | Inf ]

Dates on which the regression will be run; Dates=Inf means the entire range available will be used.

Intercept=false [ true | false ]

Include an intercept in the regression; Intercept=true means the intercept will be placed last in the matrix of explanatory variables.

Weights=[] [ Series | empty ]

Time series of regression weights on the observations in individual periods; Weights=[] means equal unit weight on all observations.

Description

This function calls the built-in lscov function.

Example

Generate random explanatory variables x and y and noise e, construct a dependent variable a, and estimate two regressions, one excluding the intercept (not included in the "true" relationship), the other including the intercept.

x = Series(qq(2020,1), rand(1000,1));
y = Series(qq(2020,1), rand(1000,1));
e = Series(qq(2020,1), 0.1*randn(1000,1));
a = 0.5*x - 0.5*y + e;

[est1, stdEst2] = regress(a, [x, y])
[est2, stdEst2] = regress(a, [x, y], intercept=true)

regress (Series)