Besides ARDL & VECM, in the Multiple OLS, I regress the dependent Y(0) time series with independent time series (19 in number) X(0) and (4 in number) X(1) variables. The residuals are serially uncorrelated at BG Serial correlation tests, the coefficients stable at 5 percent level, the CUSUM Stability Diagnostic test shows stable at 5.2 percent level, R-square value is 0.313, Adj. R-square is 0.206, F-value (p-values) is 2.916 (0.001), DW stat is 2.092. Can I accept the OLS method besides the ARDL and VECM observations?
$\begingroup$
$\endgroup$
12
-
$\begingroup$ The observations in Multiple OLS is to some extent important since it has the long-run relationships. $\endgroup$– Dr. Paritosh Chandra SinhaCommented Mar 31, 2020 at 20:35
-
$\begingroup$ What is multiple OLS? While ARDL and VECM are models, OLS is not a model but an estimation technique. Do you perhaps mean multiple regression? $\endgroup$– Richard HardyCommented Apr 1, 2020 at 6:33
-
$\begingroup$ Actually, besides the ARDL and VECM I have tried to get the long run OLS relationship as it is there. The ARDL and VECM models are fine but with reduced information since none of them use level data. Hence, I have also include the OLS estimated results. Here, the dependent Y is I(0), a few of the X are I(1) while the most are I(0). The residual in the Multiple OLS is I(0). Hence, even if the model is marginally stable in the terms of level of significance i.e. 5.2percent. Will I neglect this long run specifications totally. Making first differences violates the meaning the proxy valuables do. $\endgroup$– Dr. Paritosh Chandra SinhaCommented Apr 1, 2020 at 7:15
-
$\begingroup$ Yes, it means multiple OLS regression. $\endgroup$– Dr. Paritosh Chandra SinhaCommented Apr 1, 2020 at 7:16
-
$\begingroup$ VECM does not suffer from reduced information. After all, it has an equivalent representation as a restricted VAR model in levels. $\endgroup$– Richard HardyCommented Apr 1, 2020 at 8:17
|
Show 7 more comments