Create delay_summary function built on predict_delay_parameters

[athowes]

Can use something like:

  summarised_draws <- draws[,
    list(
      mean = mean(value, na.rm = TRUE),
      median = median(value, na.rm = TRUE),
      q2.5 = quantile(value, 0.025, na.rm = TRUE),
      q5 = quantile(value, 0.05, na.rm = TRUE),
      q20 = quantile(value, 0.2, na.rm = TRUE),
      q35 = quantile(value, 0.35, na.rm = TRUE),
      q65 = quantile(value, 0.65, na.rm = TRUE),
      q80 = quantile(value, 0.8, na.rm = TRUE),
      q95 = quantile(value, 0.95, na.rm = TRUE),
      q97.5 = quantile(value, 0.975, na.rm = TRUE)
    ),
    by = by_cols
  ]

Main questions here for me are handling of how we name the groups for the samples. I think this will have interplay with #213 and the possible output of that function.

For example do we just want numbers like 1, ... or do we want to have different columns like age, location, sex, ... and then the corresponding samples (and eventually summaries). Should mention that in issue #213.

[seabbs]

We could also just write a summary.delay_samples method

[athowes]

See here for maybe useful resource on this: https://github.com/epiforecasts/scoringutils/pull/876/files

[seabbs]

I don't think that is useful?

I think our main choice is :

1. summary method
2. aggregate method
3. our own custom s3.

Do we want to build predict_delay_parameters or just use brms functionality directly to get delay samples (i.e posterior prediction with new data).

[athowes]

I thought that you had sent the scoringutils link above with the suggestion that it'd be useful for this issue. Did you mean it'd be useful for another issue? If so, which?

[athowes]

I'm moving this to not planned for the time being as I expect this can be handled by tidybayes after some initial investigation.