Improvement Description
The old static seaborn plots look a bit clunky. Especially when plotting numeric group effects (this will probably also be an issue with the ANOVA residual plots but have not tested). See example below.
Current Behavior
Clunkiness galore.
Proposed Behavior
better handling of numeric variables. E.g., use a color gradient instead of categorical colors when numeric values are passed (need to look at the data type, so that numbers cast as str are handled appropriately).
maybe use a better visualization overall? Static is fine for diagnostics, but something like a "volatility" chart might be a good upgrade here. E.g., turn linear-mixed-effects into a pipeline, merge diagnostics + metadata into a single file, and pass to volatility to create a separate visualization as output.
Maybe make the visualizations optional?
References
Example of ugly handling of numeric group variables:
a QZV:
linear-mixed-effects.zip
screengrab:
Improvement Description The old static seaborn plots look a bit clunky. Especially when plotting numeric group effects (this will probably also be an issue with the ANOVA residual plots but have not tested). See example below.
Current Behavior Clunkiness galore.
Proposed Behavior
linear-mixed-effectsinto a pipeline, merge diagnostics + metadata into a single file, and pass tovolatilityto create a separate visualization as output.References Example of ugly handling of numeric group variables: a QZV: linear-mixed-effects.zip screengrab: