Decentralization of Eviews model

1. The negative values of AIC and SC are determined by their definitions, such as AIC = - (1 + log (2 * PI) + log (U & # 39* U / T) + 2 (K + 1) / T, PI is pi, K is the number of explanatory variables, u is the resial vector, t is the sample size. The order of ARMA model can be determined according to the correlation graph to see the truncation of autocorrelation coefficient and partial autocorrelation coefficient. The truncated point of partial autocorrelation coefficient determines the order of autoregression, while the truncated point of autocorrelation coefficient determines the order of moving average.

2. Quantitative analysis can be done, but it can be done
I do a lot of such data analysis for others

3. Landlord, VAR does not distinguish between endogenous and exogenous variables, directly regard variables as endogenous variables. Unit root test is done one by one variables, cointegration is done with variables, Johansen test is used. The choice of lag time is to choose the order with the smallest AIC value according to the AIC test results, directly compare the size, not the absolute value, and then establish var. Number eight I don't know
in addition, I also have a question to ask the landlord. How does the joint test of multivariate distribution lag come out in question 2??? How to operate in Eviews??

4. This is complicated. Stata is recommended

5. 以AR(3)-GARCH(2,1)模型为例：

首先在主窗口输入
LS RR RR-1 -2 -3
得出
Variable Coefficient Std. Error t-Statistic Prob.
RR(-1) 0.007606 0.059014 0.128883 0.8975
RR(-2) 0.058005 0.058549 0.990707 0.3227
RR(-3) 0.121110 0.058985 2.053245 0.0410

然后在点estimate 在下拉选项中选择ARCH
在命令窗口中再次输入
LS RR RR-1 -2 -3
并在ARCH出填入2，GARCH处为1，得出结果

Variance backcast: ON
GARCH = C(4) + C(5)*RESID(-1)^2 + C(6)*RESID(-2)^2 + C(7)
*GARCH(-1)

Coefficient Std. Error z-Statistic Prob.

RR(-1) 0.013392 0.056863 0.235514 0.8138
RR(-2) 0.120481 0.062146 1.938671 0.0525
RR(-3) 0.095921 0.056070 1.710743 0.0871

Variance Equation

C 0.000127 3.59E-05 3.553327 0.0004
RESID(-1)^2 -0.043907 0.029463 -1.490253 0.1362
RESID(-2)^2 0.248625 0.078855 3.152960 0.0016
GARCH(-1) 0.079769 0.211942 0.376372 0.7066

R-squared 0.003674 Mean dependent var 0.001397
Adjusted R-squared -0.017908 S.D. dependent var 0.013305
S.E. of regression 0.013423 Akaike info criterion -5.819411
Sum squared resid 0.049910 Schwarz criterion -5.729472
Log likelihood 833.3564 Durbin-Watson stat 1.974819

RR是上证综合指数的周收益，用此AR3-GARCH2，1是用残差来检验超额收益的

6. Methods / steps
press the create file button to create a new file.

as shown in the figure below, select on the top left and type in the number of observations on the top right.

the observation data in the spreadsheet and paste it in the blank space of Eviews.

after clicking, the figure as shown in the figure below will appear.

select the parameter estimation of the model in the tool list, The following screen appears

in the top blank, type the model you want to estimate (for example: GDP C consumption), Yi= β 0+ β 1 * Xi, where GDP is Yi, C stands for β 0, consumption stands for Xi
pay attention to select least squares as the estimation method.

finally, the results of univariate linear regression model are generated

7. The result is not bad. The coefficient of ECM (- 1) is - 1.02, and it is significant. It shows that the short-term error correction is significant, and it is reverse correction.

8. Now the error correction model is widely used in the paper, I am very good at Eviews 6.0 for data analysis. I can help you if you need

9. For example, if y variable is autoregressive, enter:
LS Y C Y (- 1)
enter in the command window to get AR (1)

or enter:
LS y c ar (1)

there is a slight difference between the two.

10. The negative values of AIC and SC are determined by their definitions, such as AIC = - (1 + log (2 * PI) + log (U & # 39* U / T) + 2 (K + 1) / T, PI is pi, K is the number of explanatory variables, u is the resial vector, t is the sample size. The order of ARMA model can be determined according to the correlation graph to see the truncation of autocorrelation coefficient and partial autocorrelation coefficient. The truncated point of partial autocorrelation coefficient determines the order of autoregression, while the truncated point of autocorrelation coefficient determines the order of moving average.