Page: 7 prev next

Types of Computations

Currently, we have limited ourselves to least squares estimation, and the associated computations for hypothesis tests and model assessment.

The least squares estimates,

 beta vector hat = (x^tx)^-1 x^ty,

of the unknown parameters are the most popular, and often the optimal, solution to linear problems.

Least squares estimates can be computed using algorithms based on
  • matrix inversion
  • singular value decomposition, or
  • QR decomposition