| metaFluMoDL {FluMoDL} | R Documentation |
Multivariate meta-analysis for FluMoDL objects
Description
This function runs multivariate meta-analysis (using package mvmeta)
on the first-stage coefficients of influenza (and possibly RSV) incidence proxies
for multiple 'FluMoDL' object summaries.
Usage
metaFluMoDL(summaries, par = c("H1", "H3", "B", "RSV"))
Arguments
summaries |
A list of objects of class |
par |
For which model terms (sets of coefficients) to run the meta-analysis?
Defaults to |
Value
Returns an object of class 'metaFluMoDL'. This is a list of objects of class
mvmeta, representing the results of the multivariate
random-effects meta-analysis for the sets of coefficients corresponding to each
term in argument par; they can be accessed directly using the $
operator as $proxyH1, $proxyH3 and $proxyB (and also
$proxyRSV if there were RSV terms in at least two elements of
summaries and par included "RSV" – in which case,
hasRSV() returns TRUE for objects of class 'metaFluMoDL').
However, some methods have been redefined for class 'metaFluMoDL', and do not
work the same as in simple lists. In particular: length()
returns the number of summaries (number of "studies") meta-analyzed and
names() returns the names of these summaries (if the list
in summaries argument was named).
In addition, the [[ and [ operators have been redefined for class
'metaFluMoDL', and now return the Best Linear Unbiased Predictor (BLUP)
estimates for the selected summaries ("studies"), as objects of class
summary.FluMoDL; selection can be made the usual way,
with a logical or numeric index vector, or with the summary names
(as provided by names). [ returns a list of
summary.FluMoDL objects, whereas [[ returns a single object.
The returned objects contain the string "blup" in their $type element,
to distinguish them from first-stage model summaries
or pooled
result summaries. In their $description element, they contain the
name of the respective summary ("study") if a named list had been provided
in the summaries argument of metaFluMoDL(). And finally, they contain
no $pred element, as they are not associated with a particular dataset and
cross-basis matrices (which is a prerequisite to create
crosspred objects).
The pooled coefficients (for all three or four incidence proxies) can be obtained
with function pooled(), which also returns an object of class
summary.FluMoDL that you can further use.
References
Gasparrini A, Armstrong B, Kenward MG. Multivariate meta-analysis for non-linear and other multi-parameter associations. Stat Med 2012;31(29):3821–39.