OrdinaryLeastSquares

2022년 4월 1일

Library "OrdinaryLeastSquares"
One of the most common ways to estimate the coefficients for a linear regression is to use the Ordinary Least Squares (OLS) method.
This library implements OLS in pine. This implementation can be used to fit a linear regression of multiple independent variables onto one dependent variable,
as long as the assumptions behind OLS hold.

solve_xtx_inv(x, y) Solve a linear system of equations using the Ordinary Least Squares method.
This function returns both the estimated OLS solution and a matrix that essentially measures the model stability (linear dependence between the columns of 'x').
NOTE: The latter is an intermediate step when estimating the OLS solution but is useful when calculating the covariance matrix and is returned here to save computation time
so that this step doesn't have to be calculated again when things like standard errors should be calculated.
  Parameters:
    x: The matrix containing the independent variables. Each column is regarded by the algorithm as one independent variable. The row count of 'x' and 'y' must match.
    y: The matrix containing the dependent variable. This matrix can only contain one dependent variable and can therefore only contain one column. The row count of 'x' and 'y' must match.
  Returns: Returns both the estimated OLS solution and a matrix that essentially measures the model stability (xtx_inv is equal to (X'X)^-1).

solve(x, y) Solve a linear system of equations using the Ordinary Least Squares method.
  Parameters:
    x: The matrix containing the independent variables. Each column is regarded by the algorithm as one independent variable. The row count of 'x' and 'y' must match.
    y: The matrix containing the dependent variable. This matrix can only contain one dependent variable and can therefore only contain one column. The row count of 'x' and 'y' must match.
  Returns: Returns the estimated OLS solution.

standard_errors(x, y, beta_hat, xtx_inv) Calculate the standard errors.
  Parameters:
    x: The matrix containing the independent variables. Each column is regarded by the algorithm as one independent variable. The row count of 'x' and 'y' must match.
    y: The matrix containing the dependent variable. This matrix can only contain one dependent variable and can therefore only contain one column. The row count of 'x' and 'y' must match.
    beta_hat: The Ordinary Least Squares (OLS) solution provided by solve_xtx_inv() or solve().
    xtx_inv: This is (X'X)^-1, which means we take the transpose of the X matrix, multiply that the X matrix and then take the inverse of the result.
This essentially measures the linear dependence between the columns of the X matrix.
  Returns: The standard errors.

estimate(x, beta_hat) Estimate the next step of a linear model.
  Parameters:
    x: The matrix containing the independent variables. Each column is regarded by the algorithm as one independent variable. The row count of 'x' and 'y' must match.
    beta_hat: The Ordinary Least Squares (OLS) solution provided by solve_xtx_inv() or solve().
  Returns: Returns the new estimate of Y based on the linear model.

2022년 9월 28일

릴리즈 노트

v2

Updated:
solve(x, y)
  Solve a linear system of equations using the Ordinary Least Squares method.
  Parameters:
    x: The matrix containing the independent variables. Each column is regarded by the algorithm as one independent variable. The row count of 'x' and 'y' must match.
    y: The array containing the dependent variable. The row count of 'x' and the size of 'y' must match.
  Returns: Returns the estimated OLS solution.