Estimate ARMA(1,1) using estimate: Parameter AR(1) is missing

7 views (last 30 days)
Boka Li
Boka Li on 28 Feb 2016
Answered: Andrea De Polis on 24 Jan 2019
Hi, everyone
I want to fit an ARMA(1,1) Model without constant to a time series of returns (y).
estimate(arima('ARLag',1,'MALag',1,'Constant',0),y)
The result shows like:
ARIMA(0,0,1) Model:
--------------------
Conditional Probability Distribution: Gaussian
Standard t
Parameter Value Error Statistic
----------- ----------- ------------ -----------
Constant 0 Fixed Fixed
MA{1} -0.0111097 1.40704 -0.00789581
Variance 0.000441102 8.76368e-06 50.3329
ans =
ARIMA(0,0,1) Model:
--------------------
Distribution: Name = 'Gaussian'
P: 0
D: 0
Q: 1
Constant: 0
AR: {}
SAR: {}
MA: {-0.0111097} at Lags [1]
SMA: {}
Variance: 0.000441102
Why is the AR term missing? I had tried to fit same data with same ARMA(1,1) model in other statistic software and it worked well.
Actually, I faced same problem while I was trying to fit other data with different ARMA model.
Why is this happened?
I was using Econometrics Toolbox for MATLAB R2014a.
Thanks in Advance!

Answers (2)

Hang Qian
Hang Qian on 11 Mar 2016
The problem appears unusual. I tried your codes
estimate(arima('ARLag',1,'MALag',1,'Constant',0),y)
However, the software always returns a ARIMA(1,0,1) Model.
Did you see any warning messages during estimation? If the program did return a ARIMA(0,0,1), may I suggest submitting a customer support request? Thank you.
Hang Qian

Andrea De Polis
Andrea De Polis on 24 Jan 2019
Hi, I have the same exact problem. No warning shows on dysplay, but the estimated model is an arima(0, 0, 1).
I tried with an arima(2, 0, 1) and everythng works out fine.
This problem persists on both Matlab 2018b and 2017b.
Any help?
Best,
Andrea

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!