Closed janetrbarclay closed 2 years ago
Thanks! I removed the old seaborn code. I left the StatsModel bits. I had switched to StatsModel because it gives us the ability to calculate confidence intervals for the Ar and deltaPhi (which sklearn doesn't). We're not using this now, but I'm leaving the old code in case we want to reinstate that function.
This cleans up and fixes some errors in the example workflow for the GW loss. In particular, it removes the dependencies on the StatsModels and Seaborn packages.
The test data for this workflow is the data for the Christiana. I got errors with the rgcn test data, I suspect b/c that doesn't included segments / times for which we can calculate the gw loss, though I didn't actually test that.