Skip to content

Control order of operations in convolution #42

@goldingn

Description

@goldingn

As discussed at ACEFA launch, we should be explicit in the docs about whether time-varying proportions are applied at the date of infection, or the date of observation. Given that might depend on the observation model and how the proportion was defined by the user, we should probably make it an option.

To have the proportion timeseries apply to infection dates, we would do:

expected <- convolution_matrix %*% (infections * proportion)

To have the timeseries apply to observation dates we would do:

expected <- (convolution_matrix %*% infections) * proportion

We currently don't have the parentheses which I think evaluates like the second one. I also think this should probably be the default, especially for things like DoW.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions