We are often interested in finding the boundaries of specific probability. For example, for 95% probability, we should find the boundary that includes 95% of the Gaussian function volume.
The projection of this boundary onto the - plane is the confidence ellipse. We want to find an elliptical scale factor , that extends the covariance ellipse to the confidence ellipse associated with 95% probability.
Since and represent the standard deviations of stochastically independent random variables, we can use the addition theorem for the chi-squared distribution to show that the probability associated with a confidence ellipse is given by:
For covariance ellipse , the probability associated with a covariance ellipse is
For a given probability, we can find the elliptical scale factor using
For the probability of 95%:
The properties of the confidence ellipse associated with 95% probability are:
- Ellipse center () is similar to the covariance ellipse.
- Orientation angle is similar to the covariance ellipse.
- Half-major axis length is – a scaled half-major axis of the covariance ellipse.
- Half-minor axis length is – a scaled half-minor axis of the covariance ellipse.