Multiple steps prediction using MCMC

Dear all,

Good day.
Is anyone has idea how to predict multiple steps ahead using MCMC?
based on my understanding, in MCMC, future is depend on current state.
May I know how to predict multiple steps ahead?

For example at current state t, I wish to predict t+1, t+2 and t+3?


Can you elaborate? MCMC is not really being used to predict states in our library but to create a Markov chain that resembles a posterior distribution.

Are you perhaps thinking about Markov Matrices that represent transition probabilities between states at different time steps?

Hi @ricardoV94 ,

I am investigate some speed prediction for vehicle.
For example at time t=2, I can predict the velocity for t=3.

What if I want to train a model that at time t=2, I can predict velocity for time t=3, t=4, … for multiple steps.
I wonder how can I train this model.
I have an idea of manipulating with the observed data, which means,
In original, I train a data which can observe t=2 and t=3 only.
But now I train with observed data including t=2, t=3, t=4 and so on.

But the concept still value at t3 depends on t2 and so on.
I working on time series data, so I working with sequence of data or I prefer to call it a buffer or window.
For example, I have a window size of 60, which are the past and current state at time 60, then I will predict value at t=61.
So at t=60, I can predict value at t=61.
Now I wish to predict t=61, 62 and so on, with current state of t=60.

Please let me know if my elaboration still not clear enough.