Purpose
Sets various options related to the FIT command and fitting process.
Syntax
FITOPT {N | S | I | P}
{E | V | C
{X | R}
{ITER #} {HALF #} {CONV #}
{LINEAR # | LOGLINEAR # | PLINEAR #}
{CI | NOCI} {LEVEL #}
{COVA | CORR | NOCO}
{PARA | NOPARA}
{INTERCEPT | NOINTERCEPT}
{LIST} @
Arguments and Subcommands
N, S, I, P
Controls the amount of output printed for each iteration. The choices are: N for no output; S for the total number of iterations only; I for the deviance at each iteration; P for the deviance and parameter estimates at each iteration. The default option is I.
E, V, C
Controls the output printed in the summary of results for the fitted model. The choices are: E for parameter estimates, standard errors, and standardized scores; V for the E option output plus the parameter covariance matrix; and C for the E option output plus the parameter correlation matrix. The default option is E. The CI command can be used to print the covariance or correlation matrices after a model has been fit.
X, R
Controls whether or not intrinsically aliased parameters are to be included in the parameter estimate tables produced by the FIT, MODEL, and CI commands. If the X option is chosen, parameter estimate tables will include all parameters whether or not they are aliased. If the R option is chosen, intrinsically aliased parameters will not be included in the tables. The default option is R.
ITER #
Sets the maximum number of iterations to be #. The default maximum number of iterations is 20.
HALF #
Sets the maximum number of step-halvings allowed for in an iteration to be #. The default maximum number of step-halvings is 5.
CONV #
Sets the convergence criterion to 
. The default convergence criterion is
0.0001 (that is, # = -4). Convergence is determined using the absolute change in
the deviance.
LINEAR #
Sets the subterm modified by a model formula following the FIT command to be the linear subterm of the indicated term (#). The term to which a model formula given on a FIT command applies is called the default subterm. Unless explicitly changed by this subcommand, the loglinear subterm of term 0 is the default subterm.
LOGLINEAR #
Sets the subterm modified by a model formula following the FIT command to be the loglinear subterm of the indicated term (#). The term to which a model formula given on a FIT command applies is called the default subterm. Unless explicitly changed by this subcommand, the loglinear subterm of term 0 is the default subterm.
PLINEAR #
Sets the subterm modified by a model formula following the FIT command to be the linear product subterm of the indicated term (#). The term to which a model formula given on a FIT command applies is called the default subterm. Unless explicitly changed by this subcommand, the loglinear subterm of term 0 is the default subterm.
CI
Indicates that Wald-type confidence bounds and, if requested, the parameter covariance or correlation matrix are always printed after each fit. The bounds table is not printed unless explicitly requested using the CI command. (For a definition of Wald bounds, see the CI command.) This option is particularly useful when confidence bounds are wanted for a fit with BY groups.
NOCI
Indicates that confidence bounds and covariance or correlation matrices are not to be printed after each fit.
LEVEL #
Specifies the (two-sided) confidence level. The specified level should be a number between 50 and 100. Other values will be ignored. By default, two-sided 95% confidence bounds are computed. This sets a global level which will be used for bounds produced by the CI and BOUNDS commands.
COVA
Prints the parameter covariance matrix. This is equivalent to the V subcommand. Once chosen, this option remains in effect until changed by the E, C, NOCO or CORR subcommands.
CORR
Prints the parameter correlation matrix. This is equivalent to the C subcommand. Once chosen, this option remains in effect until changed by the E, V, NOCO or COVA subcommands.
NOCO
Suppresses printing of the covariance or correlation matrix. This is equivalent to the E subcommand. Once chosen, this option remains in effect until changed by the V, C, COVA or CORR subcommands. This is the default option.
PARA
Prints the parameter estimate table. Once chosen, this option remains in effect until the NOPA subcommand is given with either the FITOPT or CI command. This is the default option.
NOPA
Cancels printing of the parameter estimate table. Choosing this option will suppress printing of the confidence bounds. This is useful if the user wishes to see only the parameter covariance or correlation matrix. Once chosen, this option remains in effect until the PARA subcommand is given with either the FITOPT or CI command.
INTERCEPT
Forces the model to always include an intercept. In AMFIT and GMBO this means that %CON will always be included in the loglinear subterm of term 0. This option will be set automatically if the user adds %CON to the default term. This command has no effect for PECAN and PEANUTS since this parameter is not estimable.
NOINTERCEPT
An intercept term will not be included in the model. This is the default condition. This option will be set automatically if the user drops %CON from the default term.
LIST
Prints a table showing the status of the fit options. If no FITOPT subcommands are given, LIST is assumed.
Remarks
Options set with a FITOPT command remain in effect during the current session or until changed by the user on a subsequent FITOPT command.
Unless explicitly modified by the user, the initial estimates used in a fit are the final estimates for the parameters from the most recent model. Thus, where convergence has not been obtained after the maximum number of iterations, the user reissues the FIT command to continue the iterations.
Step-halving is an adjustment to the Newton-Raphson step size which is taken if the likelihood fails to increase when the default step is taken. This adjustment involves moving in the same direction as the Newton-Raphson step, but only going half as far. Step-halving usually occurs when the initial estimates are very bad or when a parameter estimate is plus or minus infinity.
Examples
a) To allow the model formula on a FIT command to affect the loglinear subterm of term 1:
FITO LOGL 1 @
b) To change the
maximum number of iterations to 50, change the convergence criterion to be 
, and include the parameter covariance
matrix as part of the summary printed after the fit:
FITO ITER 50 CONV -6 COVA @
c) To print a list of the current fit options:
FITO @
d) To request that 90% confidence bounds be printed after each fit:
FITO CI LEVEL 90 @
e) To request that the parameter covariance matrix be printed after each fit without the confidence bound table:
FITO CI NOPA COVA @