(EViews10):Augmented Dickey-Fuller Test, Stationarity #adf #pp #stationarity #integration

from crunch econometrics I will show you how to perform stationary to test using the argument a dickey-fuller presidio my ear views is already launched and I have my two variables which I'll be using for this analysis I have the log of personal consumption expenditure and the log of personal disposable income here as a group data before I select these two tests to the scientific testing of the arguments a dickey-fuller I need to first plot them on a series to observe their nature so I double click on them I go to view graph and I click line on symbol okay so looking at the nature of the two graphs there were trending on board and that's an indication that both series are non stationary another way to know whether my series are non stationary is to perform the regression of log PCE unlock PDI and observe the value of my house whale and Durbin Watson side double click on both variables again I go to quick and I click estimate equation and type in my dependent variable which is locked PCE C which is a constant and log P di which is the explanatory variable I click OK the rule of thumb is that if the R squared is greater than token Watson statistic is an evidence that you have just performed a spurious regression and let's look at my R squared is point nine nine four and my WRC is point five seven zero so this is a clear indication that this regression is periods and it's poulos because both series are non stationary and one of the disadvantages of is periods regression is that the outcome or the results cannot be used for prediction or forecasting or what assist testing the outcome of explorers regression is basically useless I've been observe the plots of the two graphs and looking at the outcome of this a regression I need two subjects these two variables to the augmented dickey-fuller Thurston so the next things we do is to click on the first one the log of per capita expenditure I click on View and the unit root test dialog box opens up so first thing is that I'm going to test in level so this one remains the way it is per level and from what I've seen when I plotted the graph I will start with an intercept I will allow AIC to choose the maximum lag length from what I have here I manually inputted it Wynonna AIC will choose the optimal log line from this aids I click OK so here I have my results for the argument a Dickey fuller test used on the levels form and selecting the intercept option the output you have seen on the screen I in two parts the upper part is the unit root test itself at home of the unit root test and the lower part is a regression from the unit to test because I included an intercept in my specification if you look at the intercepts here the coefficient is points 0.5 and is statistically not significant again out of the choice of maximum of 8 labs AIC use three legs of the dependent variable which is also here three lads but the most important outcome is a new upper part of the table where we have the null hypothesis as the log of PC has a unit rule the argument says test the idea of test statistic is negative one point four seven eight because we only consider the absolute value by absolute value we don't consider the negative science and if the absolute value is lower than the critical value I cannot reject the null hypothesis that this series indeed has a unit root so from this test calm I cannot reject the null hypothesis so I click on view again I go to uni truth now let me select range and intercepts I leave every route in the where they are and I click OK selecting the trend an intercept let's look at the lower part of the table the regression art boots so here the constant term is significant the trend term is significant even though we are not considering the P values in this regression we only using the P values here but this is just lets you know that song the trained and the constant terms are significant again the most important part of these outputs on the upper part of the table where we have the null hypothesis as in log of PC has a unit root from here we can see that the ADF statistic is negative three point soon I and because I'm basing my significance level at five percents I still cannot reject the null hypothesis in this case even though it is weakly significant are ten percent bomb rigged economy did the null hypothesis that in Lombok PC has a unit root at 5 percent level so in this case I cannot read L the null hypothesis so having confirmed that in level PCE is non-stationary let me not check the first difference I click on intercept for the first difference I live a cocky with a maximum of eight legs and I click OK now I have the art again in the upper parts of my table is the most important parts where see how my null hypothesis now as the difference of the log of PC having a unit root and my ADF test statistic is negative three point two AIDS and he fell to look at the absolute value is higher than the 5% critical value so looking at what I Xavier I have to reject the null hypothesis that D the difference of LM PC has a unit root so by rejecting the null hypothesis I can say that now the series is stationary now let me just again using train and intercept so I click on this one and I click OK still using the false difference now looking at the trend and intercept the null hypothesis remains the same the ADF test statistics is three point four five eight slightly lower than my five percent critical value three point six four so by including constant untrained i cannot reject the null hypothesis that is still not that it is still not stationary at first difference so that means the outcome of this test is that the log of PCE is stationary only with a constants not with a constant and trained so that would be the outcome of this series so i have tested using intercept untrained i have tested using only the intercepts with the intercepts i reject the null hypothesis that is not stationary by including the train and the intercept Akinori reject the null hypothesis the next thing I will do is to plot the series of the different log of per capita expenditure and look at a trend now let's look at those series the different series and say now this is a different log of per capita expenditure so we can see here died stationary it revolves around the mean of point zero zero zero one if I'm to draw a trend line straight horizontally from point zero zero one we can see that the series exhibits a mean reversion around point zero zero one so the log differenced log of per capita expenditure is stationary at first difference having said all that it's important for me to run you through some things that you need to know when conducting stationarity testing eviews if he is in a time series data is essentially test for stationarity it stationary series simply means it is it has the constant mean a constant variance and constant covariance in other words the series is time invariance however if that is not the case then the series is non stationary also in terms of its analyses you can use the word non stationary you can use your neutrals on random walk they are all synonyms also when you regress to series together you guess what we call the spurious regression how do you know that you have is first regression when the value of your R squared is greater than iow Watson statistics is a confirmation come block the streets on the graph to visualize the train then go ahead to perform your stationarity test by difference in the variable several chests are bound we have the augmented dickey-fuller the finished beer on the DFG LS and so on and so forth the null hypothesis is always that the series has the unit would reject the null hypothesis if the absolute value of the computer talks artistic exceeds the interpolated critical value of decaf Allah or McKinnon the preferred benchmark for signals level is 5 percent compared to it attained or 1% the difference between the ADF test and the D F test is that the algometer dickey-fuller are there the large difference of the dependent variable to take care of possible serial correlation in the error terms and lastly always plot the graph of the different series to visualize is nature thank you for watching subscribe for more videos from crunch econometrics

Channel: CrunchEconometrix

Views: 43,600

Rating: 4.9295154 out of 5

Keywords: Augmented Dickey-Fuller, Akaike criterion, critical value, first difference, Durbin-Watson, MacKinnon, nonstationary, Phillips-Perron, random walk, spurious regression, stationary, tau statistic, tren

Id: ovpHuz6YMLc

Length: 11min 23sec (683 seconds)

Published: Thu Feb 22 2018