Allow filters on participant characteristics

[brittsikora]

Filter by subgroup, (sex, median age split, and active vs. placebo when unblinded)

[jwildfire]

Adding participant-level filters on the primary page is possible (see http://graphics.rhoworld.com/tools/aeexplorer/example4/), but complicated. The most robust way to handle it is to have two separate data inputs for the chart. One event-level data set for AEs and one participant-level data set for demographics. Would break backwards compatibility, but will consider for version 3.

Other options are possible, but are going to involve complex settings for filters, which I want to avoid.

[jwildfire]

I'm leaning towards adding a type parameter to filter specifications instead of requiring multiple files. Options for settings.filters[x].type would be:

• 'event' for filters related to the AEs themselves, like event Severity (this is the currently supported functionality)
• 'participant' for filters related to the participants, like participant Gender (not currently supported)

The new filters specification would look like this:

filters:  [   
  {'value_col': 'AESER'
  ,'label': 'Serious?'
  ,'type':'event'} //new in v3.0.0
  ,{'value_col': 'GENDER'
  ,'label': 'Gender'
  ,'type':'participant'}] //new in v3.0.0

I think code changes would be isolated to the following files:

• /src/chart/AETable/util/prepareData.js - Changes to the algorithm creating chart.filtered_data (this would represent the bulk of the updates)
• /src/chart/AETable/controls/custom/init.js - Add a label to each filter showing the type
• /src/chart/defaultSettings.js - update filter defaults to include type

I'll start on a branch off of refactor with this approach and see how it goes.

[jwildfire]

Seems like approach above will work well. Initial commit branch v3-filters. Needs some more clean-up and testing, but here's a working example.

[jwildfire]

merged in to refactor will be released in v3