feat: implement automated reporting and visualization layer - #4040
feat: implement automated reporting and visualization layer#4040ayushman1210 wants to merge 18 commits into
Conversation
divine7022
left a comment
There was a problem hiding this comment.
solid work! thanks for putting this together
the bigger picture, ryt now this is plotting layer. and 4037 is validation report, and the job of a validation report is to answer two questions "is the model good enough ?" and "where does it fail ?". a stack of pretty panels doesn't answer either. the upgrades I am trying to asking are what turn this from "here are some plots" into "here's whether the model passes, and where it doesn't" none of it is hard and you have already got the pieces in.
and consolidate one off plot_time_series into these, it is not touched here. either fold it in or spin quick follow up
Co-authored-by: Akash B V <[email protected]>
- Add per-site faceting and scorecards to timeseries and residual plots - Standardize plot styles (lowercase axes, dynamic units, 12pt base_size) - Implement linear trendlines for residual plots - Incorporate obvs_sd into overlap/coverage logic - Remove redundant stop() calls and ensure terminating newlines
There was a problem hiding this comment.
thanks for the updates!,
please check all the ci failing, same as I dropped in #4041
update the changes suggested in slack discussion here or may be as followup
…ies plot & wide metric format
… lm fit in residual plot, and standardize 90% interval
|
this may not be caused by this PR, but according to CI error, may need to add to .github/workflows/ci.yml but please confirm that this is the correct solution |
divine7022
left a comment
There was a problem hiding this comment.
thanks for solid work on this, pretty close.
droped couple of inline
then it's good to merge
Description
Here's a breakdown of what I worked on:
metric_timeseries_plot.R,metric_scatter_plot.R, andmetric_residual_plot.Rfunctions so they can easily consume thedata.frameordata.tableoutputs straight from therun_benchmark()pipeline. I also took the opportunity to clean up theggplot2aesthetics (better themes, removing hardcoded date limitations, adding reference lines, etc.).Validation_report.qmdtemplate ininst/reports/. This generates a nice HTML scorecard (ILAMB-style) that compiles all our computed metrics alongside the plots.generate_validation_report.R, which acts as a wrapper to render the Quarto template automatically at the end of a benchmark run.test-visualization.Rto make sure the plots render correctly and the wrapper function handles missing templates gracefully.Types of changes
Fixes #4037