Subject specific transition matrices (HMM)

[saltwater-tensor]

Hi all, Is there a possibility of estimating subject specific transition matrices for a group level HMM model?

One possibility is: We can empirically estimate the transition matrix after using the fine_tuning function for the model.

We could use some form of Bayesian rules to use the group level transition matrix as a prior and then update the empirically estimated transition matrix.

---

Is there a better way to do this that already exists in code?

[scho97]

We have a function that takes inferred state time courses to estimate subject-level transition probability matrices. Perhaps you can check out the osl_dynamics.analysis.modes.calc_trans_prob_matrix function and see if this is what you want.

[saltwater-tensor]

Thanks, this is similar to what I have implemented.

Furthermore, once state time courses are derived using the fine-tuning function, incorporating group-level transition matrices, there is no need for Bayesian updating of the empirically derived transition matrix.

Thank you for your help.