Mitchell O'Hara-Wild

This may need an option, such as `drop` or `trim` which removes structurally redundant nodes in the hierarchy.

MRE: ``` r library(fabletools) library(tsibble) tibble::tribble( ~ grp_a, ~ grp_b, ~ time, ~ y, "a", NA, 1, 1, "b", "x", 1, 2, "b", "y", 1, 3 ) %>% as_tsibble(c(grp_a, grp_b),...

It's stored in the model object, but this isn't intended for users. If needed I can expose it in the user accessible methods for the model. Are you trying to...

> With `devtools::revdep_check()` I have the following error: > *Error: 'revdep_check' is not an object exported from 'namespace: devtools'* The devtools package doesn't provide reverse dependency checking functions. You get...

This unexpected behaviour is due to how the scaling term for `MASE` is computed. The `by` argument is used to group up the data passed to the accuracy functions, and...

Compute MASE with accuracy separately for each series, then compute the mean from the result. ``` r library(fpp3) #> ── Attaching packages ──────────────────────────────────────────── fpp3 0.4.0 ── #> ✓ tibble 3.1.6...

I think an alternative approach that would work here is reconsidering how accuracy measures are specified. Instead of always reducing the output into a single unit summary, it could be...

Right. I'll try to rework the `accuracy()` interface to compute accuracy measures in the two stage approach mentioned above.