Hello to all,
How to compute the conditional probability of a time series having multiple dimensions X belongs to Rd, in a way P(X=xi|X=xj) is the likelihood of observing xj given xi and also the reverse of it P(X=xj|X=xi), provided X fits its density estimate, PX(x)?
Thanks
Hi,
Mathematically, P(X=xi|X=xj) = 0 if xi is different of xj. Unless I'm missing something, I think you should introduce time indices in your variable X
Best regards.
Ouzzani Fares Thanks for your answer. Although, I understand that the conditional probability could be zero for some relationships, it can't be for all if I compute it between consecutive data instances or separated few samples. I can actually solve the above using HMM using states instead P(X=xi|Si) but I don't want to profile the distribution.
I also understand that I have to include constraints while solving the above but isn't x~(PX(x)) sufficient enough?
I hope I am not getting very ambitious.
thanks
- Badges
- Science topic
- Similar topics
- Statistics
- Time Series