Need to check whether these small differences are avoidable within arma_onestage() and arma_multistage(), or if they instead need to be sandpapered over in multistage_rcpp().
Instead of returning a true zero, some of the variances (!!) are tiny, tiny negative numbers, which results in NaNs when using SE(). This has to be avoided.
Need to check whether these small differences are avoidable within
arma_onestage()
andarma_multistage()
, or if they instead need to be sandpapered over inmultistage_rcpp()
.Instead of returning a true zero, some of the variances (!!) are tiny, tiny negative numbers, which results in
NaN
s when usingSE()
. This has to be avoided.