This function uses the data available in a FluMoDL object
to generate predictions (in the form of crosspred objects) for a
summary.FluMoDL object.
addPredictions(s, m)An object of class summary.FluMoDL (normally holding BLUP or
pooled estimates, i.e. s$type will equal "blup" or "pooled")
for which predictions will be generated.
An object of class FluMoDL, which provides the original (untransformed) predictor data to create the predictions.
The function returns the summary.FluMoDL object s,
with predictions added (as element $pred)
Creating a cross-basis matrix (to use as covariate in a Distributed-Lag Nonlinear Model) transforms and scales the original predictor. Interpreting the model coefficients requires revisiting the cross-basis matrix and backtransforming to the original predictor, in order to generate predicted effect estimates for specific values of predictor and lag.
For this reason, summary.FluMoDL objects created from a
multivariate meta-analysis, containing pooled or BLUP coefficients,
do not contain predictions (their $pred element is NULL)
because they have no reference to an original predictor.
This is what addPredictions() does: it uses the cross-basis matrices from a
FluMoDL object m to calculate predictions with the coefficients in
the summary.FluMoDL object s. It provides the necessary
"context" in which to interpret the model coefficients.