Gretl Manual: Gnu Regression, Econometrics and Time-series Library | ||
---|---|---|
Prev | Chapter 8. Nonlinear least squares | Next |
The NLS estimation procedure is an iterative process.
Iteration is terminated when a convergence criterion is met or
when a set maximum number of iterations is reached, whichever
comes first. The maximum number of iterations is
100*(k+1) when analytical derivatives are
given and 200*(k+1) when numerical
derivatives are used, where k
denotes the
number of parameters being estimated. The convergence criterion
is that the relative error in the sum of squares, and/or the
relative error between the the coefficient vector and the
solution, is estimated to be no larger than some small value.
This "small value" is by default the machine
precision to the power 3/4, but it can be set with the
genr command using the special variable
toler
. For example
genr toler = .0001
will relax the tolerance to 0.0001.