Forecast corrector #68
Conversation
|
Synthetic data? :) |
add coverage to pull_request_tests
reVX/utilities/forecasts.py
Outdated
| Average MEA for all sites | ||
| """ | ||
| agg_mae = (np.abs(np.nansum(fcsts - actuals, axis=1)).sum() | ||
| / np.nansum(np.nanmax(actuals, axis=0) * 8760)) |
There was a problem hiding this comment.
8760 hard coded? Can't you just use len()?
There was a problem hiding this comment.
whoops, took this from Greg, good catch
| Returns | ||
| ------- | ||
| agg_mae : float | ||
| Aggregate MAE for all sites |
There was a problem hiding this comment.
What is this? The sum of the MAE for all sites?
There was a problem hiding this comment.
As I remember yes. Look at the over all MAE for all sites vs the average MAE across sites
| fcsts : ndarray | ||
| Bias corrected and blended forecasts | ||
| """ | ||
| fcsts = cls.bias_correct_fcst(actuals, fcsts) |
There was a problem hiding this comment.
I think you should include in both the method doc strings and the class doc string the exact steps taken to manipulate the forecast data. This will help with transparent documentation propagated through to the docs.
reVX/utilities/forecasts.py
Outdated
| ave_mae = np.nanmean(np.abs(fcsts - actuals).sum(axis=0) | ||
| / (actuals.max(axis=0) * time_steps)) | ||
| site_mae = np.mean(np.abs(fcsts - actuals), axis=0) | ||
| site_rel_mae = site_mae / (np.max(actuals, axis=0) * time_steps) |
There was a problem hiding this comment.
wait do we actually need to divide by the number of steps anymore?
Forecast corrector
@grantbuster Thoughts on data to use for tests?