The expectation and variance of aggregated predictions is just a sum if the predictions are (conditionally) independent. This function computes the DSS for a matrix of observations and a matrix of predictions where the columns are to be summed according to a given factor.
dssAggregate(observed, pred, psi, groups)
a matrix of DSS values
a numeric matrix of observed counts.
a numeric matrix of predicted counts.
a numeric vector or matrix of overdispersion parameters such that
pred * (1 + pred/exp(psi))
is the prediction's variance.
Alternatively, psi = NULL
indicated Poisson predictions.
a factor variable of length ncol(observed)
indicating
which columns should be aggregated.