Skip to content

Use as_forecast_sample converter vs scoring wrapper #1

@seabbs

Description

@seabbs

score_census <- function(census_fc, truth) {

Hi @CyGei

This looks really nice! I realise this is early days but wanted flag that we are keen people use adapters vs wrappers when working with scoringutils as much as possible (i.e. when it makes sense for their use case). We are still thinking about what docs for this should look like but as an example here are the EpiNow2 and epinowcast adapters.

https://github.com/epiforecasts/EpiNow2/blob/main/R/as_forecast_sample.R

and

https://github.com/epinowcast/epinowcast/blob/e4987506f230a506a88010a33ad9d13d7da9b6c7/R/model-validation.R#L27

The reason we think this is the way to go is that it lets your users have a very flexible time scoring and reduces the dev burden on package developers (they don't have to keep in step with our features i.e. if we add a new scoring transform or aggregation approach).

Something to note though is that this doesn't make it harder for package devs to specify different defaults to ours so any views on that very welcome.

No pressure to make any changes here FYI!

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions