Correlate CV errors with errors between structured and unstructured data

[wlangera]

Ideally, the CV errors would be able to identify species with biggest difference from expected value according to the structured dataset (e.g. behind the purple line).

We can look how CV errors correlate with errors between structured and unstructured data.
Some considerations:

• ABV data might not be optimal to test this due to oversampling in rare habitats?
• CV errors are not able to capture this since they know nothing about the structured data. Extreme values might be more dependent on dataset eveness for example. Remake figure but with eveness values or other values in "Trends in error" section
• Might not be consistent across different indicators?