Aggregation error in R inference algorithm

[saraloo]

There is something wrong in the R classical inference when fitting data is weekly (and simulation data is daily, as always).
At first glance it looks like the simulation model output is fitting terribly to the data (below, fitting data are red points, compared to simulation in grey).

However, it appears as though there is a bug in aggregation somewhere during the likelihood calculations, because multiplying the ground truth data by 7 (or 'equivalently' dividing the simulation output by 7) looks like a great fit.

Digging a little there is a bug in the aggregation function in the inference R package where the simulation model output is filtered to only the dates that are present in the fitting data (effectively choosing only ONE day of simulated output to compare to WEEKLY fitting data).

Testing a fix for this now and will PR when I confirm that it works...

[jcblemai]

very good catch