Skip to content

Visualization As Validation

Jonathan Bloedow edited this page Nov 21, 2024 · 3 revisions

Introduction

One of the things we've been exploring in the new cholera model is using visualizations of inputs as a quick way of validating that the data ingested and potentially tuned by the model at startup is what the modeler expected. This is obviously not something one would do when running at scale, but when doing model development or running for the first time with new data inputs, it can be a very useful way to smoke test the data.

Obviously this is not the same as a quantitative check on the data, but it can still help give the modeler a chance to quickly confirm that the extents and patterns are as expected. And since LASER is a spatial model, in many cases our inputs as well as our outputs are 2D, spanning by time and space (nodes/patches).

Examples

Some examples of these input visualizations are:

Populations

image

Environmental Suitability

image

Seasonality Contact Scale Factor

image

Environmental WASH Factor

image

Other inputs which can be visualized are:

  • Initial Age Structure
  • Expected Age at Death
  • Migration Rates
  • Initial Immunity
  • Scheduled Intervention by Time and Place

Notes

  • There is a runtime option to display these individually to the screen or in aggregate to a pdf.
  • These are inputs, not outputs. The topic of visualization of model outputs is separate.
  • These are read-only. It's natural to look at these and start to think about integrating them into a GUI and making them editable. That might be a great idea but is not scoped or scheduled work at this time.
Clone this wiki locally