Page 75 - Azerbaijan State University of Economics
P. 75

Murad Yusifov: Modelling the inflationary processes and forecasting:an application of ARIMA,
                                                   SARIMA models


                                           0.005700     0.9400  0.613619  0.4364  0.002944  0.9569
                    Heteroskedasticity
                    Test: White
                                           0.795490     0.6697  0.747056  0.7972  2.527891  0.4684
                    Normality test              0,3129             0,1102              0,426

                 So, let`s look at the diagnostic test results of SARIMA (1, 0, 1) (0, 0, 1) x12:


                               Table.6. SARIMA (1, 0, 1)(0,0,1)x12 model diagnostic test results.
                                              Testlər
                                                                          F-statistika  P-qiymət
                             Breusch-Godfrey Serial Correlation LM Test:   0.870505     0.4237

                                   Heteroskedasticity Test: ARCH           1.784194     0.1862

                                   Heteroskedasticity Test: White          2.064974     0.0201

                                           Normality test                    0,578

                 If we cast a glance at the comparative statistic loss functions, such as RMSE,MAE,MAPE and

               TIC the same indicator were at least on ARIMA(2,0,1) model [4].

                Table.7. Statistic loss functions of the models.
                Tests               ARIMA(2,0,1)     ARIMA(6,0,1)  ARIMA(5,0,4)  SARIMA(1,0,1)(0,0,1)x12

                RMSE                     0,4111           0,549          0,5518                0,4865
                MAE                      0,3418          0,4711          0,4893                0,395
                MAPE                     0,3406          0,4694          0,4873                0,3931
                TIC                     0,00205          0,0027          0,0027                0,0024
                Bias proportion          0,075            0,041          0,087                 0,039


               While analyzing the statistic loss functions RMSE,MAE,MAPE and TIC  the figures were the

               least were at least on ARIMA(2,0,1).













                                                             75
   70   71   72   73   74   75   76   77   78   79   80