One way to do this that seems simple to me and also of public health relevance would be to make forecasts correspond to binned values of hospital admissions in the same example we're using for quantile, median, and mean forecasts based on forecasts of hospital admissions. This would be consistent with what was done for the old flusight forecasts if the bins were fairly narrow, and would be consistent with ideas that have been thrown out for getting to categorical levels of "intensity" like "low", "moderate", "high", and "very high" incidence as categorical measures of severity.
If we go this route, we could get to the bin probabilities from the quantile forecasts using distfromq::make_p_fn. As a reference/starting point if we decide to do this, distfromq::make_r_fn is being used for roughly similar purposes here.
One way to do this that seems simple to me and also of public health relevance would be to make forecasts correspond to binned values of hospital admissions in the same example we're using for quantile, median, and mean forecasts based on forecasts of hospital admissions. This would be consistent with what was done for the old flusight forecasts if the bins were fairly narrow, and would be consistent with ideas that have been thrown out for getting to categorical levels of "intensity" like "low", "moderate", "high", and "very high" incidence as categorical measures of severity.
If we go this route, we could get to the bin probabilities from the quantile forecasts using
distfromq::make_p_fn. As a reference/starting point if we decide to do this,distfromq::make_r_fnis being used for roughly similar purposes here.