Several observed and potential errors are linked to the retrospective intervals for sibling regression with covariates (and other models!). This requires some discussion to determine a strategy for fixing.
Issues
with short data sets, the default of at least 15 years of data for the retrospective may not be met, creating an error on the "FC Plot" and "FC Table" tabs (and on the Compare Tab)
the slider for the number of years includes impossible values:
slider starts at 5, but complex sib reg retro (all retro?) requires at least 10 obs. Don't seem to have the same min obs for other model forms (e.g., Naive and ARIMA run with 5 obs to start the retro -> but is that sensible?)
slider goes to 35, but that may not be applicable for many input files
there is currently no user guidance in the tooltip under which conditions to avoid the retrospective option
the slider range for retrospective should respond to the year range picked for the analysis
Temporary Fixes implemented so far (in the testing version for the next release)
changed the lower bound on the slide to 10 (in explore tab and compare tab)
Added brief note to tooltip that the retrospective interval may be inappropriate for short time series (few values to generate interval) and will cause an error if selected value exceeds (years - 1) for the oldest age class.
Proposed Fixes
make this dynamics:
Step 1: link upper end of slider to max available value (need some rule of thumb like "need at least x retrospective values to calculate an interval")
Step 2: make the whole retro option dependent on whether or not certain criteria are met (if not, the option doesn't show up).
Several observed and potential errors are linked to the retrospective intervals for sibling regression with covariates (and other models!). This requires some discussion to determine a strategy for fixing.
Issues
Temporary Fixes implemented so far (in the testing version for the next release)
Proposed Fixes