I am searching for a statistical test that tells wether a finite-alphabet time series is a Markov process of a given order or not.
Asked
Active
Viewed 27 times
1
-
just for clarification (I have no answer anyway): for "order" of the process you mean the number of states and not the number of previous states in the process's memory, right? – mau Sep 05 '13 at 11:44
-
I actually meant the memory of the process. – user93236 Sep 05 '13 at 12:56
-
but in this case every (finite) sequence may be seen as the result of a suitable Markow process, with its order equal to the length of the sequence :-) – mau Sep 05 '13 at 13:24
-
Assume that you fix (or upper-bound) the order beforehand. – user93236 Sep 05 '13 at 13:42