When visualizing Column Pair Trends, don't display rows/columns without any scores #832
Description
Problem Description
In Issue 815, we are updating the Quality Report to only consider pairwise trends that are significant in the real data. But the visualization for the Column Pair Trends property currently shows gray boxes for the insignificant correlations. This can make the visualization hard to parse:
Expected behavior
- The top heatmap (similarity score): We should remove any rows/columns that we can. If the entire row/column is grayed out, we can remove it.
- In the bottom 2 heatmaps (numerical correlations): The real data should have gray boxes if the correlation is not significant. Then for both heatmaps, we should also remove the entire row/column if it's gray. This means that the real and synthetic data should show the same columns (be in sync).
Additional Edge Cases
In some datasets, we may find that there are no significant column pair trends at all. In in this case, we shouldn't show any of the graphs at all.
In other datasets, there may be no significant numerical correlations. In this case, we can still show the top heatmap (overall similarity score) but we shouldn't show the bottom 2 heatmaps (for numerical correlations).