For my university project I examine the effect of American tariffs and customs and other import duties on the European export (time-series data, quarterly, 89 observations for all variables). So far, I have run a regression analysis of my data (picture 1), determined that my data is nonstationary at Level but stationary at First Difference. Also, according to DW Stat (0.124752
Dear Sir,
First, the stability of the series is examined. The unit root test is performed (Dickey-Fuller). If the series is not stationary, the first difference is taken.
After the series is stationary, the ARIMA model is searched.
Eview program
Quick/Series Statistics/Unit Root/Serial name
Answer (Continued)
Quick/Estimation equation
Variable c MA(1) MA(2) AR(1) AR(2)
Şenol Çelik thank you, I enclose ARIMA output and ARMA output for you to see if I am following corretly
Hello,
AR(1) and AR(2) are significant, but MA(1) and MA(2) are not significant.
AR(2) model is suitable.
Şenol Çelik Ok, great, thank you! And shall I use the AR for all my variables or just for one (the dependent)? And how can I incroporate the variable with AR into regression equation in EViews? Every time when I estimate the equation as y c x1 x2 dummy ar(2) the output makes my independent variables not significant according to P-value.
If the dummy ar(2) is making your independent variables not to be significant, then you need to try with ar of different orders may be ar(2) is not adequate.
Emmanuel Alphonsus Akpan thanks for comment! I ended up running VAR because no combination of AR and MA would work for my data. They remained autocorrelated even though I tried to add AR, MA, Y(-1), lagged values, logs, etc. Do you think VAR is a good solution? So far the analysis shows no autocorrelation or heteroscedasticity, so I am quite happy, but still not sure..
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Applied Statistics
- Survey Design
- Questionnaire