TABFIT <table>,<name of model>,L
estimates the parameters of a log-linear model for frequency data
given as a multiway table (F table) or
estimates the parameters of an ANOVA model for data
given as a multiway table (X table).
Both types of TABFIT models are special cases of GLM (generalized linear
models) and treated here by the same program.
The type of the multiway table (F or X) is given as the last parameter
of the TABLE definition for <table>. It specifies whether a log-linear
model (with the log link) or a classical ANOVA model (with the identity
link) is estimated.
TABFIT <table>,<name of model>,L
Parameter L is optional and gives the first line for the results in
the edit field.
The model to be fitted has to be defined by a MODEL specification typed
on any two consecutive lines in the edit field as follows:
MODEL <name of model>
<list of terms to be fitted>
<table> has to be written in the edit field and it has to be defined
by a TABLE specification. The structure of <table> is that of tables
produced by the TAB operation (See TAB?). Thus, multiway tables
computed by TAB can be used as input for TABFIT without any changes.
Each term in the model may consist either of one (main effect) or
several (interaction between factors) factors.
The factors are notated by the initials of the classifier names
in the table.
The model is given on two consecutive edit lines in the form:
MODEL <name of the model>
<list of the terms to be fitted>
The terms to be fitted are separated by a '+' operator. If the term is
an interaction term of two or several classifiers, the factors must be
separated by '.'.For example, A.B represents an interaction of A and B.
Also a '*' operator can be used:
For example, A*B=A+B+A.B
A*B*C=A+B+C+A.B+A.C+B.C+A.B.C
The user can remove any existing term from the model by operator '-':
For example, A*B*C-A.B.C=A+B+C+A.B+A.C+B.C
The constant term is given as '1'.
Thus a typical model definition is
MODEL M1
1+S+B*K
where S,B and K are initials of classifiers of the multiway table.
The extent of results given by TABFIT is controlled by the RESULTS
specification.
If RESULTS=0, only minimal information (i.e. deviance, degrees of freedom
and structure of the model) is printed.
Otherwise also the estimates of the parameters and their standard errors
are given.
Furthermore, if a value >70 (like RESULTS=100) is given, the fitted
values are displayed as a new multiway table having the same
structure as the original table.
In all cases, the design matrix generated by TABFIT is saved as a matrix
file XTAB.M and the covariance estimates of parameters as a matrix file
PCOV.M on the current data disk.
T = More information on table management
S = More information on statistical analysis
Help System (web edition)