Compute Expectation of Autoregressive Moving-Average Process from Its Definition 

This example explores an ARMA process with initial values. It constructs process values in terms of the sequence of innovations and uses enhanced support of random processes in Expectation to compute the mean and covariance of the process slices. Furthermore, stationary time series process is reinterpreted as time series process with random initial values.

Define autoregressive moving-average process values via its defining relation as a function of driving white noise process values .

In[1]:=
Click for copyable input
X
Out[1]=

Process values for ARMA(2,1) process.

In[2]:=
Click for copyable input
X

Compute mean of process values for , given past process values, and given zero values of innovations in the past.

In[3]:=
Click for copyable input
X
In[4]:=
Click for copyable input
X
Out[4]=

Compare to the values of the process mean function.

In[5]:=
Click for copyable input
X
Out[5]=

Use Expectation to compute covariance function of process values under the same conditions.

In[6]:=
Click for copyable input
X
Out[6]//Short=

Covariance function for a particular value of process parameters.

In[7]:=
Click for copyable input
X
Out[7]//MatrixForm=

Compare computed covariance matrix to values of CovarianceFunction.

In[8]:=
Click for copyable input
X
Out[8]=

Compute mean and covariance of process values assuming a Gaussian joint distribution for past values and past innovations.

In[9]:=
Click for copyable input
X
In[10]:=
Click for copyable input
X
In[11]:=
Click for copyable input
X
Out[11]=
In[12]:=
Click for copyable input
X
Out[12]//Short=

Weak stationarity condition implies that the mean values should be equal and the covariance matrix values across subdiagonals should be the same. This determines parameters of the joint distribution of past values.

In[13]:=
Click for copyable input
X
Out[13]=

Compare to the covariance function of weakly stationary ARMA(2,1) process.

In[14]:=
Click for copyable input
X
Out[14]=