Closed athowes closed 3 years ago
~5% of rows produced using softmax based on samples using inla.posterior.sample from (some subset of) the models result in NaN. Note that moving to Besag from BYM2 has lowered this to a consistent ~1% of rows.
softmax
inla.posterior.sample
NaN
~5% of rows produced using
softmax
based on samples usinginla.posterior.sample
from (some subset of) the models result inNaN
. Note that moving to Besag from BYM2 has lowered this to a consistent ~1% of rows.