System Identification Toolbox

No Covariance

Evaluating and visualizing the uncertainty of the estimated models is a very important aspect of system identification. Handling, and translating covariance information takes a major part of the time in many of the routines of the System Identification Toolbox. For example, in n4sid, calculating the Cramer-Rao bound (which in this case is used and an indication of the covariance properties) takes much longer than estimating the actual model. In d2c and c2d, most of the time is spent on covariance handling. If you build models that are of a preliminary nature, and you would like to speed up the calculations, you can add the property name/property value pair 'Covariance'/'None' to the list of arguments in most relevant routines. This will prevent covariance calculations and set a flag not to spend time on this in future use of the model. This flag can also be set in the model at any time by

• Model.cov = 'no'
  

The Estimated Parameter Covariance Matrix

nk and InputDelay