Initial Values for regARIMA Model Estimation
estimate
uses fmincon
from Optimization Toolbox™ to minimize the negative loglikelihood objective function.
fmincon
requires initial (i.e., starting) values to begin the
optimization process.
If you want to specify your own initial values, then use name-value pair arguments.
For example, to specify 0.1 for the initial value of a nonseasonal AR coefficient of the
error model, pass the name-value pair argument 'AR0',0.1
into
estimate
.
By default, estimate
generates initial values using standard time
series techniques. If you partially specify initial values (that is, specify initial
values for some parameters), estimate
honors the initial values that
you set, and generates default initial values for the remaining parameters.
estimate
enforces stability and invertibility for all seasonal and
nonseasonal AR and MA lag operator polynomials of the error model. When you specify
initial values for the AR and MA coefficients, it is possible that
estimate
cannot find initial values for the remaining
coefficients that satisfy stability and invertibility. In this case,
estimate
honors your initial values, and sets the remaining
initial coefficient values to 0.
The way estimate
generates default initial values depends on the model.
If the model contains a regression component and intercept, then
estimate
performs ordinary least squares (OLS).estimate
uses the estimates forBeta0
andIntercept0
. Then,estimate
infers the unconditional disturbances using the regression model.estimate
uses the inferred unconditional disturbances and the ARIMA error model to gather the other initial values.If the model does not contain a regression component and an intercept, then the unconditional disturbance series is the response series.
estimate
uses the unconditional disturbances and the ARIMA error model to gather the other initial values.
This table summarizes the techniques that estimate
uses to gather the remaining initial values.
Technique to Generate Initial Values | ||
---|---|---|
Parameter | Error Model Does Not Contain MA Terms | Error Model Contains MA Terms |
AR | OLS | Solve the Yule-Walker equations [1]. |
MA | N/A | Solve the Yule-Walker equations [1]. |
Variance | Population variance of OLS residuals | Variance of inferred innovation process (using initial MA coefficients) |
References
[1] Box, G. E. P., G. M. Jenkins, and G. C. Reinsel. Time Series Analysis: Forecasting and Control. 3rd ed. Englewood Cliffs, NJ: Prentice Hall, 1994.