I've been working with survival and risk models lately and was wondering how similar statistical models in general could be incorporated into an uzulang, or used within existing uzulangs, since they are ultimately functions of time.
For example: with an accelerated failure time model you have a survival curve where the probability of an event not happening is either accelerated or decelerated by the effect of each input variable. In the case of medicine this would be something like age or white blood cell count.
Could one set a pattern where the values of these probability inputs vary over a given number of cycles and change the probability of an event being triggered over time? While we do already have straightforward "probability of event" controls with functions like sometimes(), something like this might be helpful for generative patterns where probabilities of one event happening could be used to change probability of a different event happening. I might also be overthinking something that can be solved with existing functions so I'd be interested in any patch ideas for Strudel/Tidalcycles that do the same thing.
More info on Survival analysis:
More on Accelerated failure time models:
Really good book on statistical models and statistical learning in general:
Without going too much into detail of these particular models (because I am not very familiar with these kind of models):
Probabilities of events happening which influences the events sounds a lot like a Markov chain - Wikipedia which uses an additional state variable to describe the system.
Within this family, there are more simpler ones like AR (auto regressive) and MA (moving average). I'm actually not sure if this a prime example of a time-dependent, but non-uzu signal because at every time step, the value depends on all the previous ones.