philliplab / tsic Goto Github PK

Too many hacks to compensate for flaws in the algorithm.

Just design the algorithm better and most of the hack will become unnecessary

Downstream analyses require estimates of the probability that infection occurred on each day. Thus compute a vector of probabilities with elements for each day and each element being the estimate of the probability that infection occured on the corresponding day.

This should be performed by computing the integral of the aggregate function from the start of the day to the start of the next day.

There are performance concerns: First prove that they are real.

There are concerns about the integrals converging properly: First prove that this is a real issue, then use a bunch of little hacks to check if it was computed correctly and then change the starting conditions of the integration function accordingly.

For now put this code in interpretation_processing and refactor later.

Given dx hist and ordered list of tests (by window period) return trimmed dx hist with only most informative + and - for each sample. In almost all cases, this will return only one test per sample date.

This function is doing so much more than it was originally intended to do, split it up.