Multivariate (Vector) adaptive AR estimation base on a multidimensional
Kalman filer algorithm. A standard VAR model (A0=I) is implemented. The
state vector is defined as X=(A1|A2...|Ap) and x=vec(X')

[x,e,Kalman,Q2] = mvaar(y,p,UC,mode,Kalman)

The standard MVAR model is defined as:

y(n)-A1(n)*y(n-1)-...-Ap(n)*y(n-p)=e(n)

The dimension of y(n) equals s

Input Parameters:

y			Observed data or signal
p			prescribed maximum model order (default 1)
UC			update coefficient	(default 0.001)
mode	 	update method of the process noise covariance matrix 0...4 ^
correspond to S0...S4 (default 0)

Output Parameters

e			prediction error of dimension s
x			state vector of dimension s*s*p
Q2			measurement noise covariance matrix of dimension s x s