Dataset schema: metrics error

[javi-gv94]

Some communties (e.g. QfO, CAMEO) consider an error/confidence interval for the metrics apart from the actual value of the metric (stored in the assessment dataset schema), which is an important part of the visualization. How should we keep track of those errors?

• Create a new attribute in the dataset schema to store the metric error value.
• Consider the metric error as a 'new' metric, which would mean adding more testactions and datasets to the data model structure.

[jmfernandez]

Could you provide one or two example links to existing entries requiring it, please?

[javi-gv94]

Examples from Quest for Orthologs

For instance: Species Tree discordance benchmark has error for one of the metrics, while Swisstree has error for both metrics

[jmfernandez]

Links are working, but they are showing nothing in a cold load. Could you attach screenshots to the issue, please?

[javi-gv94]

Sorry, I was updating the previous post. Here an example for one of the benchmarks in QfO:

[jmfernandez]

I guess you are referring to this:

Metric values can be as structured as you need. In this concrete case, what you can do is storing as metric value a JSON object with two keys, metric value itself and metric value error. Is it feasible?

[javi-gv94]

Yes that could be the solution. We can encode both values in the data-uri attribute.

[jmfernandez]

Then I'm keeping this issue open until you have tried that approach

[jmfernandez]

Applied changes on commit fddca4987b77624732d68e4692b57c1f69c52ac2 which allow having complex, structured inline values