Adding lateral flow test intervention code. #194
Conversation
There was a problem hiding this comment.
Given the size of this change, please can you perform some performance profiling analysis (we really need to add this a reg test as well). There are a few places which looks like it may be adding overhead (even if the feature is turned off), so we need to understand that.
tests/test_interventions.py
Outdated
| model.one_time_step() | ||
| results = model.one_time_step_results() | ||
|
|
||
| np.testing.assert_equal( results[ "n_lateral_flow_tests" ] > 0, True, "Expected lateral flow tests not found." ) |
There was a problem hiding this comment.
Can this have a better estimate of the number of tests that we would expect? Non-zero is a pretty low bar to check.
| """ | ||
| end_time = test_params[ "end_time" ] | ||
|
|
||
| params = utils.get_params_swig() |
There was a problem hiding this comment.
You should be able to replace these 4 lines now with simply
model=Model(params = test_params)
And add Model to the from model import statement at the top.
tests/test_interventions.py
Outdated
|
|
||
| # check the sensitivity at the peak. | ||
| df_test[ "symptomatic" ] = ( df_test["time_symptomatic"] > 0 ) | ||
| true_pos = df_test[ ( df_test[ "time_since_inf" ] == 3 ) & ( df_test[ "symptomatic" ] == True ) & ( df_test[ "lateral_flow_status" ] == 1 ) & ( df_test[ "test_sensitive_inf" ] == True ) ].groupby( [ "time_since_inf" ] ).size() |
There was a problem hiding this comment.
Why 3? Would be good if this was not hard-coded but linked back to the code in the model.
| @@ -1219,12 +1366,21 @@ void intervention_on_traced( | |||
|
|
|||
| parameters *params = model->params; | |||
|
|
|||
| int lateral_flow_test = params->lateral_flow_test_on_traced && gsl_ran_bernoulli( rng, params->lateral_flow_test_fraction ); | |||
There was a problem hiding this comment.
If params->lateral_flow_test_on_traced=FALSE does the gsl_ran_bernoulli() get called?
Probably best to move to inside the if statement below.
There was a problem hiding this comment.
No, C does lazy boolean evaluation.
And I think it makes sense to order lateral flow tests even if quarantine_on_traced is off. Unless you don't want that to happen? Or maybe you meant move the rng into the lateral flow if? But I think whether or not the quarantine exclusion happens should depend on if they accepted the lateral flow test.
|
|
||
| } | ||
|
|
||
| /***************************************************************************************** |
There was a problem hiding this comment.
This function does not seem to be used (unless I've missed it's usage). Please remove.
There was a problem hiding this comment.
It's used in the python. I'm also fine removing it explicitly from the python, but I want to be able to at least call it manually in py each time step to understand the sensitivity.
| } | ||
| intervention_quarantine_until( model, indiv, NULL, time_event, TRUE, NULL, model->time, 1 ); |
There was a problem hiding this comment.
Shouldn't this remain inside the if block?
Also, if the individual does not quarantine, then the household members should not quarantine either.
There was a problem hiding this comment.
I want to create the trace token whether or not quarantine_compliance_positive is true so tracing can still happen even if they don't quarantine (since the upload fraction is a separate parameter).
And both their quarantine and the household will be based on the time_event, so their quarantining should remain the same, no?
Will clean up spacing soon, just wanted to get this out for PR.
* Breaking out lfa test helpers. * Gating computation on performing LFA. * Condensing redundancy.
* Making test cleanup event driven. * Additional tests for number of tests performed.
|
For some reason I can't respond to some comments, so adding it here:
|
Includes base code to test on symptoms and tracing, as well as a number of related parameters.
Note that the specific parameters (g/b) are currently untuned and not ready for production use.
This could have a few different alternate implementations, open to tweaks as desired, especially with regard to how the infectiousness curve is created and housed. Fine deleting the extra variables from the one_time_step_results too, but they were useful to have.
I have also added a bit of extra cleanup and parameters. If you need this separated out, I can.