FIT Statement

SAS/INSIGHT Statements

FIT Statement

FIT variable-list < = effects-list >
          < / < FREQ=variable > < WEIGHT=variable >
          < LABEL=variable > < NOINT >
          < RESP=response > < BINOM=variable >
          < OFFSET=variable > < LINK=link >
          < POWER=value > < NOEXACT > < FISHER >

< QUASI > < SCALE=scale > < CONSTANT=value > >; FIT variable-list < = effects-list >

< / < FREQ=variable > < WEIGHT=variable >

< LABEL=variable > < NOINT >

< RESP=response > < BINOM=variable >

< OFFSET=variable > < LINK=link >

< POWER=value > < NOEXACT > < FISHER >

< QUASI > < SCALE=scale > < CONSTANT=value > >;

The FIT statement creates a fit analysis.

You must specify at least one Y variable. You can follow the Y variables with an equal sign (=) and a list of model effects, including simple, crossed, and nested effects:

   Y = X
   Y = X1 * X2
   Y = X( A B )

If you do not specify an effects list, a model with only an intercept term (mean) is fit.

You can use the FREQ=, WEIGHT=, and LABEL= options to assign Freq, Weight, and Label variables. FIT statement options default to fit classical linear models, but you can set them to fit the generalized linear model.

NOINT: Use the NOINT option to fit a model without an intercept term.
RESP=response: For response distribution, choose NORMAL, INVGAUSS, GAMMA, POISSON, or BINOMIAL. By default, RESP= is NORMAL.
BINOM=variable: Use the BINOM= option to specify a Binomial variable when RESP=Binomial. When RESP is not Binomial, the BINOM= option is not used.
OFFSET=variable: Use the OFFSET= option to specify an Offset variable.
LINK=link: For link function, choose CANONICAL, IDENTITY, LOG, LOGIT, PROBIT, CLOGCLOG, or POWER. By default, LINK= is CANONICAL.
POWER=value: Use the POWER= option to set a value for the POWER link function. If LINK= is not set to POWER, the POWER= option is not used.
NOEXACT: Use the NOEXACT option to fit a linear model without using exact distributions for the test statistics.
FISHER: Use the FISHER option to use Fisher's scoring method in the maximum-likelihood estimation for the regression parameters.
QUASI: If overdispersion is present in the model, you can use the QUASI option to fit the generalized linear model using the quasi-likelihood functions. To use the QUASI option, you must also set the SCALE= option to a scale other than MLE.
SCALE=scale: For scale, choose MLE, DEVIANCE, PEARSON, or CONSTANT. By default, SCALE= is MLE (maximum-likelihood estimate).
CONSTANT=value: Use the CONSTANT= option to set a constant value when SCALE=CONSTANT. If SCALE= is not set to CONSTANT, the CONSTANT= option is not used.

Chapter Contents
Previous
Next
Top