Sir, may the heavens smile upon you for spreading your precious knowledge. Your lordship gained my eternal gratitude and subscription. Greetings from Turkey.
helpful video. what I still don't get is whether the "adf test" only tests for a stochastic trend but not a deterministic trend. In other words, it tests for the existence for a random walk but not the existence of a trend, as the final decision is based on the t-statistic of the coefficient of the lagged dependent variable in the regression table below the actual unit root table, while it ignores the coefficient for the time trend and the coefficient for the intercept. Like when you tested for the stationarity of the first difference, the trend was significant, but doesn't a significant trend make the series non-stationary?
Hi! Thank you for the videos! I just want to understand one thing. Pls, I am using ARDL, and in my model, I included the square (quadratic form) of one variable to check if there exists a non-linear impact. But whenever I try it gives me a singular matrix. but I really need to use the quadratic form in the model. please I need advice on this and how to go about it. thank you. Please am using Eviews
Hi... I have a question about the ADF test. I wanted to know that is it better to always start the test from the trend and intercept option? and then if the prob at the top of the page showed the existence of unit root test, we look up for its reason in the prob of trend. and if the trend was significant, then we go for first or second difference, and if the series became stationary at I(1) or I(2) is it necessary to repeat the test for intercept only or none? please answer my question because i have been looking up every where to find the best guideline for it but i haven't found yet. thank you.
Thanks! In this case it is obvious that we reject the null hypothesis. But how would you interpret the results if you get a t-statistic that is between 2 levels of confidence? For example: how would you interpret the test, if you get a Test statistic of -3.5839
Thank you, I have one question, is it statistically correct to regress explanatory variable which is stationary at level on a first differenced dependent variable ? or else do we have to take the first difference of the stationary explanatory variable whatsoever before running the regression ?
I have a question.. U showed that the LNEX(-1), C and @Trend are significant. Does that mean it is rejecting he Null Hypothesis "LNEX has a Unit Root".. Nice explaination by the way. Thanks for your effort indeed.
Please, if your variables are sationary at I(1) and one of the variable is in I(0). when doing cointegration do will drop the one that integrated at level 0. when doing the dynamic regression do will include the variable will droped when we are doing cointegration? It this procedure correct? Thanks (
Hi Sayed i have two question for you and i need your reply as soon as possible. i have convert all data to Log then test the data using four unit root test namely:the Augmented Dickey-Fuller (ADF), the Detruded Dickey-Fuller (DF-GLS), the Phillips-Perron (PP), and the Kwiatkowski, Phillips, Schmidt, and Shin (KPSS) test .Two test show me that some variables are stationary at the first difference and the other two test show me some variables are stationary at the second difference. So in this case can i apply Johansen Test of Cointegration? The second question. when i apply Johannes Test of Cointegration i have to use the non stationary data after convert them to log or before convert them to Log
Hello, I apply the ADF test to a series, and when I select the option levels, the series I reject the null hypothesis but , same as you, when I select the option of first difference, I cannot reject the null, I made the same process as you of genere that series and the graph show me a stacionary process, But what it means??.. thanks , regrets from Colombia, your videos have help me a lot, thanks.
Hi, this is Beag Fayaz requesting all the experts to whom it may concern. I have run the ADF test for a time series data which show me the non-stationarity at level when considering the constant term in the data. but on the other hand when i utilize the same with considering trend and intercept, the results show the stationarity at level. Again when i use "none", i.e., neither trend nor intercept, the results come out showing non stationarity in the series. Now here is my query regarding the ADF tets: 1. Two out of the three models, i.e., (a) intercept and (b) none, show me the non-stationarity in the series when utilising the test at level. What should i conclude? 2. While testing the ADF at first difference, all the three models, i.e., (i) intercept, (ii) intercept and trend, and (iii) none, show me the series is staionary. Again, What should i conclude?
Hi. I'm sure your problem has been solved already but I'd like to take a shot at it. Well, based on the video's explanation is, first you have to draw a graph from your data. And I'm not sure, but I think you need to transform your data to a logarithm form. Then you interpret your time-series data from the graph. If it's like the video than it's a time series with trend and constant because it follows a random walk with drift. If it rises constantly and doesn't go down, it's likely to be time series with constant only. And if it's centered around zero value, it may be a time series with no constant and no trend (none). After that, maybe you can apply the correct ADF test to your data.
Oh, It' s very helpful lesson! Thank you. Could you help me with the following issue: time series contain a structural break, maybe two. What kind of unit root test should I do?
+LondonPhD, Is Andrew and Zivot test in the Eviews 8 or 9 ? And if this test shows I(1) and breakpoint, could I use Granger causality test or VECM? Or if this test does not identify I(1), I could not apply Granger test and other co-integration test? Thank you in advance. I apologize for any mistakes.
Yes, you can but this test let's you know where the break is and it accounts for it when testing for unit root. You can run a VECM and then granger causality if your series are I(1). If it does not say I(1) but says I(0), you should do level VAR without differencing and do granger causality the same way. hope this helps
SO, WHAT CAN WE DO WITH SOME SERIES HAS UNIT ROOT AT LEVEL WITH TREND AND INTERCEPT AND HAS NOT UNIT ROOTWITH TREND AND INTERCEPT AT FIRST DIFFERENCE ?ALSO WITH THE SERIES HAS UNIT ROOT WIT THE INTERCPT BUT NOT WITH OTHERS BUT BECOME HASNOT UNIT ROOT AFTER FIRST DIFFERCEING, IN THIS CASE CAN I RUN JOHANSEN COINTEGRATION TEST OR NOT? PLEASE COULD YOU PROVIDE ME THE ANSWER?
thanks for the video, actually I am doing my analysis with r and the results were confusing to interpret in the sense that I had forgotten that we need to take the values in absolute values
Hi thank you for the useful video I have one question is ... ADF augmented dickey test is only for time series right .. if it is panel data I need to use other method ? Thank you
dear kareem. if your data stationery at different level you still can continue with cointegration test. i suggest you use ARDL suggest by pesaran et al. (2001)....
congratulations on your most recent success!may the Good Lord continue to open more fruitful doors of achievements in your life. We are all so proud of you,keep up the spirit!
If you mean a unit root test that accounts for structural breaks, look up the perron unit root breakpoint test (1 structural break) or Lee-Strazicich test (2 breaks).
Sir, may the heavens smile upon you for spreading your precious knowledge. Your lordship gained my eternal gratitude and subscription. Greetings from Turkey.
What is the name of the software or application the person is using??? Plz tell me
Thank you. A most appreciated video, that both provides intuition for the unit root testing strategy and interpretation.
this is the most informative tutorial i have heard so far many thanks
Thanks a lot. I am teaching myself how to use E-views for data analysis and your video has come in handy.
Sweet, I was having problems with the constant term in models. Good and clear explanation.
Thank a lot Doc, it was very helpful for my masters thesis.
Thanks, this was very useful with my econometrics assignment.
Fantastic video@LondonPhD, just a quick one, which software is this?
Thank you so much for this, it was really helpful for me!
>it was really helpful
sent me ur email address
helpful video. what I still don't get is whether the "adf test" only tests for a stochastic trend but not a deterministic trend. In other words, it tests for the existence for a random walk but not the existence of a trend, as the final decision is based on the t-statistic of the coefficient of the lagged dependent variable in the regression table below the actual unit root table, while it ignores the coefficient for the time trend and the coefficient for the intercept. Like when you tested for the stationarity of the first difference, the trend was significant, but doesn't a significant trend make the series non-stationary?
Hi! Thank you for the videos! I just want to understand one thing. Pls, I am using ARDL, and in my model, I included the square (quadratic form) of one variable to check if there exists a non-linear impact. But whenever I try it gives me a singular matrix. but I really need to use the quadratic form in the model. please I need advice on this and how to go about it. thank you. Please am using Eviews
Thanks Doc, your vdo's are awesome!
Hi... I have a question about the ADF test. I wanted to know that is it better to always start the test from the trend and intercept option? and then if the prob at the top of the page showed the existence of unit root test, we look up for its reason in the prob of trend. and if the trend was significant, then we go for first or second difference, and if the series became stationary at I(1) or I(2) is it necessary to repeat the test for intercept only or none? please answer my question because i have been looking up every where to find the best guideline for it but i haven't found yet.
thank you.
what do I do if I cant find a variable (with only 8 observations) stationary at level, 1st difference and 2nd difference using the ADF test in Eviews?
Thank you,
if one of the three models of ADF test (intercept or none) does not bring the stationarity in second differnced......what is the decision?
Thanks! In this case it is obvious that we reject the null hypothesis.
But how would you interpret the results if you get a t-statistic that is between 2 levels of confidence?
For example: how would you interpret the test, if you get a Test statistic of -3.5839
Thank you, I have one question, is it statistically correct to regress explanatory variable which is stationary at level on a first differenced dependent variable ? or else do we have to take the first difference of the stationary explanatory variable whatsoever before running the regression ?
Thanks a lot for this excellent video
I have a question.. U showed that the LNEX(-1), C and @Trend are significant. Does that mean it is rejecting he Null Hypothesis "LNEX has a Unit Root".. Nice explaination by the way. Thanks for your effort indeed.
Please, if your variables are sationary at I(1) and one of the variable is in I(0). when doing cointegration do will drop the one that integrated at level 0. when doing the dynamic regression do will include the variable will droped when we are doing cointegration? It this procedure correct? Thanks
(
Hi Sayed
i have two question for you and i need your reply as soon as possible.
i have convert all data to Log then test the data using four unit root test namely:the Augmented Dickey-Fuller (ADF), the Detruded Dickey-Fuller (DF-GLS), the Phillips-Perron (PP), and the Kwiatkowski, Phillips, Schmidt, and Shin (KPSS) test .Two test show me that some variables are stationary at the first difference and the other two test show me some variables are stationary at the second difference. So in this case can i apply Johansen Test of Cointegration?
The second question.
when i apply Johannes Test of Cointegration i have to use the non stationary data after convert them to log or before convert them to Log
Thanks, this video is very informative!!
Thank you very much. You are my hero. Are there any prediction time-series data? like neural network or exponential smoothing method
Hello, I apply the ADF test to a series, and when I select the option levels, the series I reject the null hypothesis but , same as you, when I select the option of first difference, I cannot reject the null, I made the same process as you of genere that series and the graph show me a stacionary process, But what it means??.. thanks , regrets from Colombia, your videos have help me a lot, thanks.
thank you so much for the help you clearly explained the concepts for me
thanks, this was very useful
Hi, thank you. How can I convert negative variables into natural log in eviews?
Hi, this is Beag Fayaz requesting all the experts to whom it may concern. I have run the ADF test for a time series data which show me the non-stationarity at level when considering the constant term in the data. but on the other hand when i utilize the same with considering trend and intercept, the results show the stationarity at level. Again when i use "none", i.e., neither trend nor intercept, the results come out showing non stationarity in the series.
Now here is my query regarding the ADF tets:
1. Two out of the three models, i.e., (a) intercept and (b) none, show me the non-stationarity in the series when utilising the test at level. What should i conclude?
2. While testing the ADF at first difference, all the three models, i.e., (i) intercept, (ii) intercept and trend, and (iii) none, show me the series is staionary. Again, What should i conclude?
Hi. I'm sure your problem has been solved already but I'd like to take a shot at it. Well, based on the video's explanation is, first you have to draw a graph from your data. And I'm not sure, but I think you need to transform your data to a logarithm form.
Then you interpret your time-series data from the graph. If it's like the video than it's a time series with trend and constant because it follows a random walk with drift. If it rises constantly and doesn't go down, it's likely to be time series with constant only. And if it's centered around zero value, it may be a time series with no constant and no trend (none).
After that, maybe you can apply the correct ADF test to your data.
I realized that you are Uzbek. In which University do you teach? Your econometrics explanations are very useful. I found them interesting!
very usefull in fact ! Thanks a lot !
Oh, It' s very helpful lesson! Thank you. Could you help me with the following issue: time series contain a structural break, maybe two. What kind of unit root test should I do?
+Зюзя Селезнева try Andrew-Zivot test
+LondonPhD, Is Andrew and Zivot test in the Eviews 8 or 9 ? And if this test shows I(1) and breakpoint, could I use Granger causality test or VECM? Or if this test does not identify I(1), I could not apply Granger test and other co-integration test? Thank you in advance. I apologize for any mistakes.
Yes, you can but this test let's you know where the break is and it accounts for it when testing for unit root. You can run a VECM and then granger causality if your series are I(1). If it does not say I(1) but says I(0), you should do level VAR without differencing and do granger causality the same way. hope this helps
What statistic program did you use in this video?
what the common second difference in eviews ? Y C X1 X2
im runinng ADF test in eview, if my model i select second difference, the serieal model only become I(0) , is it okay with second difference ?
Buen video. Muchas gracias.
what if the constant remains insignificant through all differences (including level) with intercept as well as trend and intercept?
What is the name of the software or application the person is using??? Plz tell me
videos on var, vec and johansen tests are available on youtube from others. if you cannot find them, then I will upload some
would be better
thank you very much , u made my day buddy
SO, WHAT CAN WE DO WITH SOME SERIES HAS UNIT ROOT AT LEVEL WITH TREND AND INTERCEPT AND HAS NOT UNIT ROOTWITH TREND AND INTERCEPT AT FIRST DIFFERENCE ?ALSO WITH THE SERIES HAS UNIT ROOT WIT THE INTERCPT BUT NOT WITH OTHERS BUT BECOME HASNOT UNIT ROOT AFTER FIRST DIFFERCEING, IN THIS CASE CAN I RUN JOHANSEN COINTEGRATION TEST OR NOT? PLEASE COULD YOU PROVIDE ME THE ANSWER?
thanks for the video, actually I am doing my analysis with r and the results were confusing to interpret in the sense that I had forgotten that we need to take the values in absolute values
Hi thank you for the useful video
I have one question is ... ADF augmented dickey test is only for time series right .. if it is panel data I need to use other method ?
Thank you
Yeah it's only for time series data.
hello sir i would like apply simple regression on my data after 1 st difference. please suggest me about how to apply.
it was useful,thanks so much
Thank you very much.
Thank you very much Dr
very helpful...Thanks!
thank you, helped a lot!
Thanks a lot so clear !
Hi, thanks this was quiet helpful. May you kindly enlarge the format when typing in the command window next time please.
dear kareem. if your data stationery at different level you still can continue with cointegration test. i suggest you use ARDL suggest by pesaran et al. (2001)....
Thank you Dr
thank you very much
thank you so much :)
THANK YOU!
congratulations on your most recent success!may the Good Lord continue to open more fruitful doors of achievements in your life.
We are all so proud of you,keep up the spirit!
George Dapel Thank you very much bro!!
Nice Explanation
good lecture :-)
Thanks a lot
the tutorial is good and understandable. Thanks for it.
may I know how can I download free version of eviews?
Make one on structural breaks
If you mean a unit root test that accounts for structural breaks, look up the perron unit root breakpoint test (1 structural break) or Lee-Strazicich test (2 breaks).
Thanks.. but this quality of the video and sound was so bad it gave me a headache!
good one but why dont you reply the queries in comment
thanks
thanx :)
do it in stata/
Thanks a lot!
thanks