matlab和eviews估计ARMA模型的系数不同？这是为什么

Question

数据为股票的收益率，为下列数字：

0.005731807

-0.001109208

0.000468482

-0.005140637

0.001559812

0.000469942

0.00039557

-0.002078488

0.000866562

0.000569046

-0.001906342

0

-0.002729666

0

0.002035548

-0.000868325

0.000744325

0.003317902

-0.001509024

-0.000520027

0.000272429

-4.95E-05

4.95E-05

-0.001610885

-4.96E-05

-0.000471388

0.000396973

0.000421611

-0.000247985

0.000471119

-0.00057033

0.000223212

-0.000818686

-0.000645498

0.000943279

-0.000297782

0.000818686

-0.000297626

0.000297626

0.000223162

-0.000917761

0.000124071

0.00079369

0.000470956

-0.000223057

利用matlab中ARIMA和ESTIMATE函数构造ARIMA（2,0,1）模型，系数如下:其中constant代表什么？
ARIMA(2,0,1) Model:
--------------------
Conditional Probability Distribution: Gaussian

Standard t
Parameter Value Error Statistic
----------- ----------- ------------ -----------
Constant -4.63912e-06 0.000301491 -0.0153872
AR{1} -0.474955 0.162136 -2.92937
AR{2} 0.219548 0.162411 1.35181
MA{1} 0.206515 0.249437 0.827923
Variance 2.16804e-06 1.05713e-06 2.05088

为何和eviews预测结果不一样？

Dependent Variable: SERIES01
Method: Least Squares
Date: 10/31/14 Time: 21:58
Sample (adjusted): 3 45
Included observations: 43 after adjustments
Convergence achieved after 12 iterations
MA Backcast: 2

Variable Coefficient Std. Error t-Statistic Prob.

AR(1) 0.534648 0.149625 3.573255 0.0009
AR(2) 0.120683 0.133158 0.906308 0.3702
MA(1) -0.999681 0.030206 -33.09520 0.0000

R-squared 0.288671 Mean dependent var -0.000117
Adjusted R-squared 0.253105 S.D. dependent var 0.001318
S.E. of regression 0.001139 Akaike info criterion -10.64984
Sum squared resid 5.19E-05 Schwarz criterion -10.52697
Log likelihood 231.9716 Hannan-Quinn criter. -10.60453
Durbin-Watson stat 2.190904

Inverted AR Roots .71 -.17
Inverted MA Roots 1.00

Answer 1

稍有不同很正常，不必纠结
我替别人做这类的数据分析蛮多的追问

constant是啥？arma模型中没这项。