# Estimation (generalized linear mixed models)

The model building algorithm uses a doubly iterative process that consists of an inner loop and an outer loop. The following settings apply to the inner loop.

- Parameter Convergence.
- Convergence is assumed if the maximum absolute change or maximum relative change in the parameter estimates is less than the value specified, which must be non-negative. The criterion is not used if the value specified equals 0.
- Log-likelihood Convergence.
- Convergence is assumed if the absolute change or relative change in the log-likelihood function is less than the value specified, which must be non-negative. The criterion is not used if the value specified equals 0.
- Hessian Convergence.
- For the Absolute specification, convergence is assumed if a statistic based on the Hessian is less than the value specified. For the Relative specification, convergence is assumed if the statistic is less than the product of the value specified and the absolute value of the log-likelihood. The criterion is not used if the value specified equals 0.
- Maximum Fisher scoring steps.
- Specify a non-negative integer. A value of 0 specifies the Newton-Raphson
method. Values greater than 0 specify to use the Fisher scoring algorithm
up to iteration number
*n*, where*n*is the specified integer, and Newton-Raphson thereafter. - Singularity tolerance.
- This value is used as the tolerance in checking singularity. Specify a positive value.

Note: By default, Parameter Convergence is used, where the maximum Absolute change
at a tolerance of 1E-6 is checked. This setting might produce results
that differ from the results that are obtained in versions before
version 22. To reproduce results from pre-22 versions, use Relative for
the Parameter Convergence criterion and keep the default tolerance
value of 1E-6.