ARDL Estimation in EViews

autoregressive distributed like a RDL models are standard least-squares regression which include lags of both the dependent verbal and independent variables as regresses although a RDL models have been used in econometrics for many decades they have experienced a new popularity of late as a method of explaining dynamic long-run and cointegrating relationships between variables enews offers a simple easy to use AR DL estimation engine that automatically affects the appropriate number of lags of each variable to include in the estimation and provides post estimation tools to view the long run or Co integrator relationship form of the estimation and perform bounds testing to demonstrate where we use data made available by William green in his textbook econometrics analysis the data consists of a number of core to the US macroeconomic variables dating from 1950 to 2000 following one of the ARD l examples in this book we will examine the relationship between the log of real consumption and the log of real gdp real consumption data is held in the series real cons real GDP is held in the series real GDP if we look at a graph of real Khans we can see that it is as we would expect upward trending similarly real GDP is also up or trending to estimate an i LDL model we click on quick estimate equation and then use the method drop-down to select a IDL in the dynamic specification box we enter the name of our dependent variable the log of real consumption followed by the name of the dynamic aggressor the log of real GDP we're strictly used to select the optimal number of lags of each of these variables by using the automatic selection button and then for both the dependent variable and the regressor we set the maximum of likes to consider to eight following greens example we'll use a constant and a set of costly W variables as fixed regresses we can use Eve uses ecstatics pan function to generate the entire set of dummy variables switching to the option tab allows us to choose the criteria used for model selection will select the hanging Quinn criteria clicking ok produces the estimation results the top part in estimation output shows summary information about the estimation including the name the dependent variable the date and time of estimation the number of included observations the number of different models evaluated and perhaps most importantly the final selected model in this case an ard l5 one model the middle part shows the individual coefficient estimates for the selected model we can see that each of them regresses are statistically significant apart from the quarterly dummy variables bottom part displays summary statistics based on the final information including the r-squared which is very high for our estimation and the information criteria to judge how strongly the selected five one model is preferred over other models we click on view model selection suddenly criteria graph here we can see the hand Queen criteria for each of the top 20 different ard L models a lower value is better the ARD l5 one model appears we strongly prefer to the other models it's also notable that the top three models all include five legs of the dependent bear ball we can view the cointegrating an online relationship defined by these ard L estimates by clicking on view coefficient Diagnostics cointegrating and long-run form in our case we can see from the long-run coefficient estimate of the log of real GDP that the effect of GDP on consumption is essentially no long term light effects the coefficient is very close to 1 to further test the existence of a long-run relationship between wheel consumption and wheel GDP we can perform the Bands test of presser on Shannon Smith by clicking on view coefficient Diagnostics bounce test in our case the F statistic of the bounce s is below each of the reported critical values as we failed to reject the null hypothesis that there is no long-run relationship you

Channel: IHSEViews

Views: 126,910

Rating: 4.6175709 out of 5

Keywords: EViews 9, ARDL, Bounds Test, cointegration

Id: 542uZVNHvpY

Length: 4min 58sec (298 seconds)

Published: Fri Mar 20 2015