We have an example notebook demonstrating the estimation of time-varying transition matrices, but we don't have any corresponding tests. There isn't necessarily any special code in that example that needs to be unit tested, but the example itself makes for a good functional/integration-like test of many things at once.
Also, working on this is a good way to start discussing the topic and testing alternative approaches.
Regarding the exact test, let's use something simpler than the example notebook: a model with constant emissions and known initial state probabilities. This will allow us to focus exclusively on the estimation of the parameters that drive the time-varying dynamics.
We have an example notebook demonstrating the estimation of time-varying transition matrices, but we don't have any corresponding tests. There isn't necessarily any special code in that example that needs to be unit tested, but the example itself makes for a good functional/integration-like test of many things at once.
Also, working on this is a good way to start discussing the topic and testing alternative approaches.
Regarding the exact test, let's use something simpler than the example notebook: a model with constant emissions and known initial state probabilities. This will allow us to focus exclusively on the estimation of the parameters that drive the time-varying dynamics.