model.frame.mixmeta function

Extract Model Frame and Design Matrix from mixmeta Objects

These method functions return the model frame and design matrix for meta-analytical models represented in objects of class "mixmeta".

## S3 method for class 'mixmeta'
model.frame(formula, ...)

## S3 method for class 'mixmeta'
model.matrix(object, ...)

Arguments

• object, formula: an object of class "mixmeta".
• ...: further arguments passed to or from other methods.

Details

The model frame is produced by mixmeta when fitting the meta-analytical model, and stored in the mixmeta object if argument model=TRUE. Alternatively, the model frame is directly returned from a call to mixmeta with argument method="model.frame". The method function model.frame simply extracts the saved model frame if available, or otherwise evaluates a call to mixmeta when method="model.frame".

The method function model.matrix extracts the design matrix for the fixed-effects part of a fitted meta-analytical model. It first extract the model frame by calling model.frame, and then passes the call to the default method.

Note that the model frame of mixmeta models consist of terms for both the fixed and random-effects parts, the latter including also the grouping factors. This information can be used to reconstruct the proper model frame or matrix for each part.

These methods functions are similar to those provided for regression objects lm and lm.

Returns

For model.frame, a data.frame with special attributes (see the default method model.frame) and the additional class "data.frame.mixmeta".

For model.matrix, the design matrix used to fit the model.

Author(s)

Antonio Gasparrini <antonio.gasparrini@lshtm.ac.uk >

Note

The reason why these specific method functions are made available for class mixmeta, and in particular why a new class "data.frame.mixmeta" has been defined for model frames, lies in the special handling of missing values in multivariate meta-analysis models fitted with mixmeta. Methods na.omit and na.exclude for class "data.frame.mixmeta" are useful for properly accounting for missing values when fitting these models.

See Also

See the default methods model.frame and model.matrix. See na.omit and na.exclude on the handling of missing values. See mixmeta-package for an overview of the package and modelling framework.

Examples

# RUN THE MODEL AND SUMMARIZE THE RESULTS
model <- mixmeta(cbind(PD,AL) ~ pubyear, S=berkey98[5:7], data=berkey98,
  method="ml")

# MODEL FRAME
model$model
model.frame(model)
update(model, method="model.frame")
class(model.frame(model))

# MODEL MATRIX
model.matrix(model)