The function listed here are removed from KFAS.
# S3 method for KFS
deviance(object, ...)# S3 method for SSModel
subset(x, element, states, etas, series, times, ...) <- value
subset(x, ...) <- value
# S3 method for SSModel
subset(x, element, states, etas, series, times, ...)
Subset-based methods were removed as they were somewhat confusingly
named as the subset generic in base, and most likely not
that useful (compared to [<-.SSModel).
Deviance.KFS was removed as it was mostly useless. The value was not a \(-2*(logL-logL*)\) where \(L\) is the likelihood and \(L*\) is the saturated likelihood. Instead it was based on the conditional likelihood \(p(y|theta)\) i.e. it disregards the effect of hidden states. Therefore the value returned by this function did not make much sense in non-GLM setting.
From rstandard.KFS and residuals.KFS: Computation of deviance residuals.
This option was meant to be used only for the GLM comparisons, as their
generalization to other models is lacking, but in order to avoid misleading
results in non-GLM settings, these are now completely removed.