Attribute-based Distributions

[jawalonoski]

trafficstars

Distributions can be based off attribute values (the wiki was wrong):

For example:

"distribution": { "attribute" : "probability_of_fall_injury", "default" : 0.06 },

See https://github.com/synthetichealth/synthea/issues/289

[jawalonoski]

https://github.com/synthetichealth/synthea/wiki/Generic-Module-Framework%3A-Transitions#named-probabilities

[arscan]

Interesting, this is a powerful feature and enables some pretty cool system-dynamics type modeling.

[arscan]

@jawalonoski it looks like this already exists. When looking at a distributed transition, click on 'Change to named transition', and you get a view like this (on the second transition):

It appears to show up in the JSON correctly too. Am I missing something?

I should change 'Weight' to 'Default Weight', and remove hte 'not adding up to 100%' warning, as that value is expected to change over time though.

[arscan]

And I guess if any of the transitions are named, we should allow any number, instead of constraining to 0-1, right?

[jawalonoski]

Correct, if any transitions are named, you could say something like "Weights might not add up to 100% when using named transitions." Otherwise, you can use the current warning language.

[arscan]

Changed Weight to Default Weight in #181

Will leave open to handle the 'not add up to 100' case, but will set to 'low priority enhancement' as it doesn't prevent usage.