.-
help for ^unidiff^
.-

Uniform layer effect models
---------------------------

	^unidiff^ cellvar ^,^ ^r^ow^(^rowvar^)^ ^c^olumn^(^colvar^)^ ^l^ayer^(^layvar^)^
		^e^ffect^(^effectname^)^ ^p^attern^(^patternname^)^
		[^q^uasi ^d^esign^(^varlist^)^ ^s^cores^(^varlist^)^ ^ex^tra^(^varlist^)^ ^ref^cat^(^#^)^
		 ^con^straints^(^numlist^)^ ^la^mbda^(rawlog^|^rawexp^|^stdlog^|^stdexp)^
		 ^shd(log^|^exp)^ ^savee^xp^(^newvar^)^ ^savel^ambda^(^newvar^)^ ^nodet^ail
		 ^nodisprc^ ^nodispex^tra]

where effectname is one of

	^null^ | ^add^ | ^addlin^ | ^mult^

and patternname is one of

	^fi^ | ^qpm^ | ^qs^ | ^cp^ | ^ua^ | ^re^ | ^ce^ | ^rce^ | ^hrce^ | ^own1^ | ^own2^


Description
-----------

^unidiff^ estimates the null, additive, linear additive, and multiplicative
uniform layer effect models for the analysis of differences in two-way
associations, displays relevant goodness-of-fit statistics and parameter
estimates, and optionally computes several ancillary quantities of interest.
The dataset to be analyzed must include at least four variables:
    ^cellvar^ contains the observed cell frequencies that make up the three-way
            contingency table object of analysis.
    ^rowvar^  indexes (and optionally labels) the I categories of the row
            variable R.
    ^colvar^  indexes (and optionally labels) the J categories of the column
            variable C.
    ^layvar^  indexes (and optionally labels) the K categories of the layer
            variable L.


Options
-------

^row(^rowvar^)^ is always required. It specifies the name of the row variable.

^column(^colvar^)^ is always required. It specifies the name of the column
    variable.

^layer(^layvar^)^ is always required. It specifies the name of the layer variable.

^effect(^effectname^)^ is always required. It specifies the type of uniform layer
    effect model to be estimated.
    
    ^effect(null)^ estimates the null effect model.

    ^effect(add)^ estimates the additive model.
    
    ^effect(addlin)^ estimates the linear additive model.
    
    ^effect(mult)^ estimates the multiplicative model.
    
^pattern(^patternname^)^ is always required. It specifies the baseline pattern of
    association between variables R and C. Some patterns are allowed only when
    I=J.
    
    ^pattern(fi)^ specifies the "full interaction" (saturated) pattern of
    association.
    
    ^pattern(qpm)^ specifies the "quasi-perfect mobility" pattern of association.
    It is allowed only when I=J.
    
    ^pattern(qs)^ specifies the "quasi-symmetry" pattern of association. It is
    allowed only when I=J.
    
    ^pattern(cp)^ specifies the "crossing parameters" pattern of association. It
    is allowed only when I=J.
    
    ^pattern(ua)^ specifies the "uniform association" pattern of association.
    
    ^pattern(re)^ specifies the "row effects" pattern of association.
    
    ^pattern(ce)^ specifies the "column effects" pattern of association.
    
    ^pattern(rce)^ specifies the "row and column effects I" pattern of
    association.
    
    ^pattern(hrce)^ specifies the "homogeneous row and column effects I" pattern
    of association. It is allowed only when I=J.
    
    ^pattern(own1)^ specifies a user-defined pattern of association expressed by
    a single "topological" (i.e., categorical) variable.
    
    ^pattern(own2)^ specifies a user-defined pattern of association expressed by
    one or more quantitative variables.

^quasi^ requires that the "quasi-version" (i.e., with diagonal-specific
    parameters) of the selected pattern of the R-C association be applied. It
    is allowed only when I=J.
    
^design(^varlist^)^ is required if ^pattern(own1)^ or ^pattern(own2)^ is specified.
    It specifies the list of variables that express the user-defined baseline
    pattern of association between variables R and C.

^extra(^varlist^)^ specifies a list of additional variables intended to express
    particular features of the model that lie outside its standard formulation.
    
^scores(^varlist^)^ is required if ^effect(addlin)^ is specified. It specifies the
    list of variables that represent the exogenous scores assigned to the
    layers.

^refcat(^#^)^ specifies the layer to be taken as the reference category in the
    estimation of beta and phi parameters.
    
^constraints(^numlist^)^ specifies equality constraints to be imposed on the
    estimation of beta and phi parameters.
   
^lambda(rawlog^|^rawexp^|^stdlog^|^stdexp)^ displays in tabular form the total
    interaction effects estimated by the fitted model for each layer.

    ^lambda(rawlog)^ displays the raw effects in additive (logarithmic) form.

    ^lambda(rawexp)^ displays the raw effects in multiplicative (exponential)
    form.
    
    ^lambda(stdlog)^ displays the standardized effects in additive form.
    
    ^lambda(stdexp)^ displays the standardized effects in multiplicative form.
    
^shd(log^|^exp)^ displays in tabular form layer-specific structural shift
    parameters (with standard errors), structural distances (with standard
    errors), mean structural distances, and overall structural effect computed
    according to the Sobel-Hout-Duncan approach to mobility table modeling.
    These quantities are particularly relevant in the analysis of social
    mobility tables and can be computed only when I=J.
    
    ^shd(log)^ displays all the above quantities in additive (logarithmic) form.
    
    ^shd(exp)^ displays all the above quantities in multiplicative (exponential)
    form.
    
^saveexp(^newvar^)^ creates newvar containing the expected cell frequencies under
    the fitted model.
    
^savelambda(^newvar^)^ creates newvar containing the total interaction effects
    estimated by the fitted model. The effects are saved in standardized
    additive form.
    
^nodetail^ suppresses the output describing the structure of the contingency
    table object of analysis and the specification of the fitted model.

^nodisprc^ suppresses the output of the table reporting the parameter estimates
    associated with the variables that express the R-C association pattern.

^nodispextra^ suppresses the output of the table reporting the parameter
    estimates associated with the extra variables.


Examples
--------

 . ^unidiff obs, row(father) col(son) lay(country) effect(add) pattern(fi)^
 
 . ^unidiff obs, row(father) col(son) lay(country) effect(add) pattern(fi) /*^
          ^*/   constraints(1 1 2) nodetail nodisprc nodispext^

 . ^unidiff obs, row(father) col(son) lay(country) effect(add) pattern(hrce)^
 
 . ^unidiff obs, row(father) col(son) lay(country) effect(mult) pattern(qs)^
 
 . ^unidiff obs, row(father) col(son) lay(country) effect(addlin) pattern(qs) /*^
          ^*/   scores(develop socdem east asia)


Author
------
      
         Maurizio Pisati
         Department of Sociology and Social Research
         University of Trento, Italy
         maurizio.pisati@@galactica.it