Performs screening tests for fixed terms in a
REML analysis (R.W. Payne).
|Controls printed output (
||Whether to exclude higher-order interactions in the conditional models (
||Terms that must always be included in the model (no tests on these terms); default
||Saves the F tests|
||Saves the Wald tests|
||Specifies the analysis whose fixed terms are to be tested; by default this will be the most recent
VSCREEN calculates marginal and conditional tests for fixed terms in a
REML analysis. By default, these are from the recent
REML analysis. However, you can take an earlier analysis, by using the
SAVE option of
VSCREEN to specify its save structure (saved using the
SAVE parameter of the earlier
In the marginal test, the term is added to the simplest possible model. For example, the main effect of
A would be added to the null model, and the interaction
A.B would be added to a model containing only the main effects
In the conditional test, the term is added to the most complex possible model that contains no terms involving the term to be tested. For example, interaction
A.B would be added to the model containing all terms except those involving
A.B (such as the interaction
A.B.C). By default, the most complex model includes terms with more factors or variates than the term being tested. For example, the interaction
C.D.E would be included when testing
A.B. You can exclude these higher-order terms by setting option
VSCREEN will print a message to remind you that this has been done).
You can specify terms that should always be included in the model by using the
FORCED option. These terms are fitted first, and are not tested.
||presents F statistics for the terms. If denominator degrees of freedom (ddf) are available from the earlier
||presents Wald statistics for the terms. These suffer from the usual biases of Wald tests in
You can save the results of the F tests and the Wald tests, in pointers, using the
WSAVE options, respectively. The elements of the pointers are labelled by the headers of the columns used in the printed output.
An advantage of using
VSCREEN to assess the fixed model, rather than running a succession of
REML analyses with different fixed models, is that the fixed terms are assessed against identical estimates of the random variation (as in an analysis of variance). When terms are dropped from (or added to) the fixed model in a
REML analysis, the random variation will change. For example, it will increase if a term with a Wald statistics greater than its number of degrees of freedom is dropped. It may therefore be difficult to reach consistent decisions about which fixed terms are genuinely required.
Once you have used
VSCREEN to decide which terms to keep in the fixed model, you can use only those terms for prediction, by specifying them in the
MODEL option of
VSCREEN defines a weighted regression, with weight matrix given by the inverse of the unit-by-unit variance-covariance matrix (obtained using the
UVCOVARIANCE option of
VKEEP.) It then calls the
RSCREEN procedure to calculate the tests.
Action with RESTRICT
Any restriction applied to vectors used in the
REML analysis will apply also to the results from
CAPTION 'VSCREEN example',\ 'Example 5.3.6 from The Guide to Genstat, Part 2 Statistics';\ STYLE=meta,plain FACTOR [NVALUES=322; LEVELS=27] Dam & [NVALUES=322; LEVELS=18] Pup FACTOR [NVALUES=322; LEVELS=2; LABELS=!T('M','F')] Sex FACTOR [NVALUES=322; LEVELS=3; LABELS=!T('C','Low','High')] Dose VARIATE [NVALUES=322] Littersize,Weight OPEN '%GENDIR%/Examples/GuidePart2/Rats.dat'; CHANNEL=chan READ [CHANNEL=chan] Dose,Sex,Littersize,Dam,Pup,Weight; \ FREPRESENTATION=2(labels),4(levels) CLOSE chan VCOMPONENTS [FIXED=Littersize+Dose*Sex] RANDOM=Dam/Pup REML [PRINT=model,components,wald] Weight VSCREEN