Forms predictions from a GLMM analysis (R.W. Payne).
Options
PRINT = string tokens |
What to print (description, predictions, backpredictions, se, sesummary, sed, sedsummary, vcovariance); default desc, pred, back, seds |
MODEL = formula |
Indicates which model terms (fixed and/or random) are to be used in forming the predictions; default * includes all the fixed terms and relevant random terms |
OMITTERMS = formula |
Specifies terms to be excluded from the MODEL; default * i.e. none |
FACTORIAL = scalar |
Limit on the number of factors or variates in each term in the models specified by MODEL or OMITTERMS; default 3 |
PRESENTCOMBINATIONS = factors |
Lists factors for which averages should be taken across combinations that are present |
WEIGHTS = tables |
One-way tables of weights classified by factors in the model; default * |
OFFSET = scalar |
Value of offset on which to base predictions; default 0 |
NBINOMIAL = scalar |
Supplies the total number of trials to be used for prediction with a binomial distribution (providing a value n greater than one allows predictions to be made of the number of “successes” out of n, whereas the value one predicts the proportion of successes); default 1 |
PREDICTIONS = table or scalar |
To save the predictions; default * |
BACKPREDICTIONS = table or scalar |
To save back-transformed predictions; default * |
SE = table or scalar |
To save standard errors of predictions; default * |
SED = symmetric matrix |
To save standard errors of differences between predictions; default * |
VCOVARIANCE = symmetric matrix |
To save variances and covariances of predictions; default * |
GLSAVE = pointer |
Save structure from the GLMM analysis; default * uses the SAVE structure from the most recent GLMM analysis |
Parameters
CLASSIFY = vectors |
Variates and/or factors to classify table of predictions |
LEVELS = variates, scalars or texts |
To specify values of variates and/or levels of factors for which predictions are calculated |
PARALLEL = identifiers |
For each vector in the CLASSIFY list, allows you to specify another vector in the CLASSIFY list with which the values of this vector should change in parallel (you then obtain just one dimension in the table of predictions for these vectors) |
NEWFACTOR = identifiers |
Identifiers for new factors that are defined when LEVELS are specified |
Description
GLPREDICT can be used after the GLMM directive to produce predictions of the values of the response variate at particular values of the variables in the fixed or random models. By default the predictions are from the most recent GLMM analysis, but you can use another analysis by supplying its save structure using the GLSAVE option.
The parameters are the same as those of VPREDICT (which GLPREDICT uses to form the predictions). The CLASSIFY parameter specifies variates or factors that are to be included in the table of predictions, and the LEVELS parameter supplies the values at which the predictions are to be made. For a factor, you can select some or all of the levels, while for a variate you can specify any set of values. A single level or value is represented by a scalar; several levels or values must be combined into a variate (which may of course be unnamed ). Alternatively, if the factor has labels, you can use these to select the levels for prediction by setting LEVELS to a text. A missing value in the LEVELS parameter is taken to stand for all the levels of a factor, or the mean value of a variate.
The PARALLEL parameter allows you to indicate that a factor or variate should change in parallel with another factor or variate. Both of these should have the same number of values specified for it by the LEVELS parameter of GLPREDICT. The predictions are then formed for each set of corresponding values rather than for every combination of these values.
When you specify LEVELS, a new factor must be defined to classify that dimension of the table. By default this will be an unnamed factor, but you can use the NEWFACTOR parameter to give it an identifier. The EXTRA attribute of the factor is set to the name of the corresponding factor or variate in the CLASSIFY list; this will then be used to label that dimension of the table of predictions.
The prediction calculations consist of two steps. The first step is to calculate a table of fitted values. The MODEL, OMITTERMS and FACTORIAL options specify the model to use for this. The formula specified by MODEL is expanded into a list of model terms, deleting any that contain more variates of factors than the limit specified by the FACTORIAL option. Then, any terms in the formula specified by OMITTERMS are removed.
The second step averages the fitted values over the classifications that are not in the list that was supplied by the CLASSIFY parameter. The WEIGHTS option can supply one-way tables classified by any of the factors in the model. These are used to calculate the weight to be used for each fitted value when calculating the averages. Equal weights are assumed for any factor for which no table of weights has been supplied. (Note, this differs from the default in PREDICT, which uses marginal weights; see the VPREDICT option ADJUSTMENT for details.) In the averaging all the fitted values are generally used. However, if you define a list of factors using the PRESENTCOMBINATIONS option, any combination of levels of these factors that does not occur in the data will be omitted from the averaging. Where a prediction is found to be inestimable, i.e. not invariant to the model parameterization, a missing value is given.
The OFFSET option specifies the offset value to use when calculating predicted means. The default is zero.
The NBINOMIAL parameter can be used to supply the total number of trials to be used for back-transformed predictions with a binomial distribution. If you provide a value n greater than one, GLPREDICT predicts the number of “successes” out of n. The default, NBINOMIAL=1, predicts the proportion of successes.
Printed output is controlled by settings of the PRINT option with settings:
description describes the terms and standardization policies used when forming the predictions,
predictions prints the predictions,
backpredictions prints back-transformed predictions,
se prints standard errors of the predictions,
sesummary prints the minimum, average and maximum standard error,
sed prints standard errors of differences between the predictions,
sedsummary prints the minimum, average and maximum standard error of difference,
vcovariance prints the variance and covariances of the predictions.
The default is to print descriptions, predictions, back-transformed predictions, and a summary of the standard error of differences. Standard errors and standard errors of differences are printed only if the predictions themselves are printed.
You can also save the results, using the PREDICTIONS, BACKPREDICTIONS, SE, SED and VCOVARIANCE options.
Options: PRINT, MODEL, OMITTERMS, FACTORIAL, PRESENTCOMBINATIONS, WEIGHTS, OFFSET, NBINOMIAL, PREDICTIONS, BACKPREDICTIONS, SE, SED, VCOVARIANCE, GLSAVE.
Parameters: CLASSIFY, LEVELS, PARALLEL, NEWFACTOR.
See also
Procedures: GLMM, GLDISPLAY, GLKEEP, GLPERMTEST, GLPLOT, GLRTEST, GLTOBITPOISSON.
Commands for: Regression analysis.
Example
CAPTION 'GLPREDICT example',\
!t('Data from an experiment on Great Knott, Rothamsted;',\
'see West, J.S., Fitt, B.D.L., Leech, P.K., Biddulph, J.E.,',\
'Huang, Y.-J. &, Balesdent, M.-H. (2002).',\
'Effects of timing of ~italic{Leptosphaeria maculans}',\
'ascospore release and fungicide regime on phoma leaf spot',\
'and phoma stem canker development on winter oilseed rape',\
'(~italic{Brassica napus}) in southern England.',\
'Plant Pathology, 51, 454–463.'); STYLE=meta,plain
SPLOAD [PRINT=*] '%data%/GtKnott2000.gsh'
GLMM [PRINT=model,components,wald,deviance; DISTRIBUTION=binomial;\
LINK=logit; DISPERSION=*; FIXED=Cultivar*Fungicide;\
RANDOM=Block/Wholeplot] LMplants; NBINOMIAL=Nplants
GLPREDICT [PRINT=predictions,backpredictions,sed; PREDICTIONS=Predictions;\
SED=sed] Cultivar,Fungicide
PEN 11,12; SYMBOL='circle'; CSYMBOL='red','blue'; CFILL='match'; SIZE=1.25
VARIATE [VALUES=11,12] pens
DTABLE Predictions; XFACTOR=Fungicide; GROUPS=Cultivar; BAR=MEAN(sed);\
BARDESCRIPTION='average sed'; PENS=pens