Charts are generated based on codebook rather than data

[jthrilly]

Charts on the overview screen are generated on variables defined in the codebook, rather than variables used in the data.

This means that studies using roster data that does not define codebook variables will not generate any charts.

We should consider if this is the intended behavior or not. Some considerations:

• If another interface modifies a roster variable, this will create a corresponding variable in the codebook, which will cause the chart to display
• If data is static, should we present a chart for it? What value would that have?

Sample protocol for this issue is the Sunbelt 2019 roster (ask Josh for access, as it contains private data).

[jthrilly]

I thought more about this, and my current view is that variables solely inside roster data should not generate charts, because we are not manipulating them within network canvas. The overview charts should be for variables created and manipulated within the interview flow.

[wwqrd]

Seems reasonable, the overview isn't for analysis, it doesn't need to show everything.