(EViews10):Augmented Dickey-Fuller Test, Stationarity
Vložit
- čas přidán 11. 09. 2024
- This hands-on tutorial teaches how to perform the augmented Dickey-Fuller Test for stationarity in EViews. If the series are not stationary, no inferences or forecasting can be made from such regressions. Stationary is a mandatory requirement in time-series analysis. This hands-on tutorial teaches how to perform the augmented Dickey-Fuller unit root test in EViews10. Here is the link to the detailed tutorial on this topic:
cruncheconomet......
Here is the link to the ex21-1.wf1 dataset (EViews file) used for this tutorial (endeavour to have a Google account for easy accessibility): drive.google.c...
Follow up with soft-notes and updates from CrunchEconometrix:
Website: cruncheconometr...
Blog: cruncheconomet...
Forum: cruncheconometr...
Facebook: / cruncheconometrix
CZcams Custom URL: / cruncheconometrix
Stata Videos Playlist: • (Stata13):Estimate and...
EViews Videos Playlist: • (EViews10):Interpret V...
CZcams recently changed the way my content will be monetised. My channel now needs 1,000 subscribers. So it would be amazing if you show your support by both watching my videos and subscribing to my channel if you haven’t done so already. Monetising my videos allows me to invest back into the channel with some new equipment so this small gesture from you will be extremely huge for me. Many thanks for your support….CrunchEconometrix loves to teach, help me stay online.
Great tutorial! You must be a very good lecturer
@@isaacmaison1799 I'm humbled by your positive feedback, Isaac...gracias!
hi.. can you please assist me with the data? i have requested to access the file. pleas help me with this. tq
@@rozianabaharin9222 Hi Roziana, kindly bear with me. Dofiles are now available for a token payment from my website. I have put up a Community notification since August regarding this and also stated the reason. Thanks for your understanding. May I know from where (location) you are reaching me?
Hi, your Chanel is very informative, please guide about FMOLS and DOLS models too. thanks
I love how you make this simple! You are the best. God bless you
I'm encouraged by your positive feedback...deeply appreciated! 🥰🙏
I will be able to finally finish my degree and confidently go for a job interview through this channel. I am doing my final year project with so much ease. Thank you so much CrunchEconometrix. I will donate to you if i get a job.
Thanks Peo, for the encouraging feedback. Deeply appreciated! Please may I know from where (location) you are reaching me?
@@CrunchEconometrix I am in Gaborone Botswana, studying Statistics and Economics in the University of Botswana. You have been a great guardian throughout my study. God bless You! 🙏
Thank you so much for all these explanations, I am studying time series and your videos are the BEST I have found in CZcams. I follow you from Panama 💜
I'm humbled by your encouraging feedback, Bernadette. Thanks!!! 💕 💕 💕
Really helped me for my thesis, thanks for the informative content!
Thanks for the encouraging feedback, Hish, deeply appreciated! Please may I know from where (location) you are reaching me?
@@CrunchEconometrix I'm watching from the UK!
I left many comments on your tutorials but every time I watch a new one, I cannot help it I want to thank you! Happy New Year!
Thanks, EUA for your encouraging words can feedback. Wishing you the best of Year 2023!❤️
My econometrics lecturer in Buckingham university just told me about your channel....Great video. Thanks so mch!
What?!?!?! Babaniji, u're kidding right???...wow!!! thanks!!! awesome! (dancing...)
@@CrunchEconometrix Dr Gurcharan Singh....I was struggling with some concepts so he told me to check out your videos that they'll be helpful. And they've truly been. Thanks so much 👍
U just made my day end so graciously well with this feedback. Thanks Babaniji and I'll be glad if you can tell others too! 💕 😊
Thanks prof. you have delivered it. Excellent delivery.
Awwww, thanks Japheth for the encouraging feedback. Deeply appreciated! 🥰🙏
You are wonderful ma, and God Almighty increase your knowledge
Thank you so much, Sage!!!
You are awesome!
You should really write a book or create a course on Data Analysis in the future, you are the best!
Maybe one day, Sir... thanks for your encouraging words. Deeply appreciated!🙏🥰
You Video has been the best so far I wish it was r software
Pinta, I appreciate your encouraging feedback. I'm upgrading my R proficiency before I start creating the videos. Hopefully, before the year runs out.
a very interesting and simple way for learning an ADF test.
Glad you liked it...thanks!
Thnx very much, Dr. structured and straight forward
Thanks for the positive feedback, Yassin...deeply grateful!
You're a great teacher!
Thanks for the encouragement, Joshua!
easy to understand and straight-forward. thank you.
U're welcome, Yasin.
this is very useful for my research..
Glad to hear that, Sir 🙏
excellent presentation
Thanks, Mr. Hussain for the encouraging feedback. Deeply appreciated! 🙏
thank you for the explanation
You are welcome, Morris 🙏
Thank you so much... you made it so simple and clear
Glad to be of help, Veena! Please may I know from where (location) you are reaching me?
@@CrunchEconometrix Im from India
really helpful. God bless you
Glad it was helpful, Caleb!
George from Nigeria. Your videos has really helped me. This is my first time using eview. Am carrying out a research on the impact of petroleum sector revenue on economic growth. Please am a bit confused on what test to conduct .
Hi George, thanks for the encouraging feedback. Deeply appreciated. Watch my EViews videos, follow the guides and analyse your data alongside.
Your efforts are very much appreciated Dr. ^^
Thanks for the encouraging words and feedback, Rafik! 💞
Really love your videos!
Thanks Vy!😊
Thank you so much ❤
U're welcome, Sir! 💞
Thank you so much.
You're welcome! 🥰
Thanks for teaching ma
It's my pleasure, Immaculate!
God bless you.
You too, Sir Obed! 💗🙏
thanks for the tutorials Dr.
U're welcome, Elrufai...may I know where you are reaching me from?
@@CrunchEconometrix from Malaysia
@@elrufaiahmad2999 Awesome! Kindly share my YT Channel link with your Malaysian students and academic community. Thanks!
Hi Rina, I'm not quite clear about what you want to do. Is it stationarity test? If so, you have to include at least "constant"... and if it's still not stationary try it with "constant and trend".
Thanks a lot, Great job.
Thanks for the encouraging feedback, Sherzad!😍
Wooow!! Thanks alot
U're welcome 🙏 🤗
Good afternoon Dr Adeleye,
Hope this mail meets you well.
I am happy to be your student having gone through your econometrix tutorial on youtube. I need clarity on differenced series
Let's assume that a series becomes stationary at 2nd difference which means that regressing with initial values will turn out spurious results.
How do you then run regression at 2nd difference for the same series?
thanks
Akan Orok
Hi Akan, thanks for following my video tutorials. Use only I(1) series. Not I(2).
Thanks Prof for the derailed explanation. Please trend vs trend and and intercept; which is preferred for unitroot test ?
Hi Lucy, start with INTERCEPT only and if still nonstationary use INTERCEPT+TREND...works most of the time.
Hands down the clearest, most informative, and overall best eViews tutorial. You're a great teacher, thank you very much for creating this content. Liked, subscribed.
I have one question though. If I want to run a VAR on 4 variables, and my Real GDP is stationary at 1st difference, it means I need to transform my variable to, say... Real GDP Growth, to be able to run a VAR, right?
Thanks DM for the positive feedback and remark on my video...and I am also humbled by your like and subscription, deeply appreciated! You need not transform real GDP into real GDP growth unless you intend estimating a growth model.
@@CrunchEconometrix So what if my variables are non-stationary at Level, but are stationary at 1st difference? Can I still run a VAR with these raw variables? Or I
must do something else beforehand? Thank you very much for your feedback!
@@dreammatter1734 Kindly watch my videos on VAR estimation. Procedures are explained. Information is not contained in any one video so attempt to watch all. Thanks.
@@CrunchEconometrix Thank you very much for your reply :)
Good job. Very selfless of you
Thanks for the kind words and positive feedback on my videos, Dortman. Deeply appreciated! 💕 May I know from where (location) you are reaching me?
Hello professor, i found your video now. It was clear and crisp. I have one query. If any variable is passing ADF with intercept but fails without intercept what is the inference we can draw. Can we still consider the series as stationary ?
Hi Krishna, I'd advise you perform the ADF test with intercept.
Good afternoon Dr Adeleye,
Hope this mail meets you well. Thank you for your series which is very helpful to me.
I am happy to be your student having gone through your econometrix tutorial on youtube. I need clarity on differenced series.
Let's assume that a series becomes stationary at 2nd difference which means that regressing with initial values will turn out spurious results.
How do you then run regression at 2nd difference for the same series? thanks
Hi Akan, I responded to you on a different thread.
Madam, thanks so much for your videos. I am in love with them. I do not know if you have an video you did for testing unit root using KPSS?
Thanks for the positive feedback, Gabriel. Appreciated! You can check other online resources for KPSS URT, thanks.
@@CrunchEconometrix thanks so much
Am carrying out a research on the impact of petroleum sector revenue on economic growth and my results is show me that oil rent has a negative and insignificant relationship to GDP
...so, go ahead with your interpretation.
Hi there! Thank you for your video , i am a fan of your video tutorials. I have a question, in my research paper i used raw data for testing Unit root, Johansen conintegration and Granger Causalty but later i used log form for VECM estimation. The question is would there be any problem if i conduct my analysis this way? The problem is i get unit root when i use log data in one of my variable in first difference. My second question is what is the difference between Pairwise granger and VAR granger causality test ? are they same? can i use VAR granger test instead of Pairwise granger? Thanks in advance!
Hi Ali, thanks for the encouraging feedback. Deeply appreciated! It is advisable to use the form of the variables for analysis for the preliminary tests. I will also advice you read up on the causality tests to understand the one to adopt.
@@CrunchEconometrix Thanks Professor for your response. To be clear, here the term form refers to the raw data right?
Form refers to how you used the variable in your regression. Log (transformed variable) or its raw (natural) form.
i like so much the method used to deliver this course. thanks so much. how can i find other movies.
Hi Valens, thanks for the positive feedback and kind remarks about my CZcams videos. Deeply appreciated! Kindly check through the 9 Playlists. Well sorted to guide you. Please may I know from where (location) you are reaching me?
Hi, I got a question. I also use AIC for testing unit root in my model, how many lag length is required for monthly data? Is there a rule of thumb? If not how to determine the max lag on this unit root test?
Hi Farrel, kindly watch my video on "Optimal Lag Selection". Well explained. Thanks.
Thanks Dr. Your Tutorials help me a lot in understanding more about econometric analysis. Pls is it possible to find correlation analysis between one independent variable and five independent variables using time series analysis by Eviews?
Hi Musa, thanks for the positive feedback on my videos. Humbly appreciated...and yes, you can do that in EViews.
Please madam can you help me with the questions below:
1. What if at first difference my data is significant can I do the regression analysis?
2. How do I lag my data
Kindly watch my TIME SERIES videos. Detailed to guide you.
Hats off to u for a good explanation. Mam i have a question: my pakistan exchange rate data series from 1987 to 2019 is stationary at second difference while the other variables are stationary at 1st difference. So what should i do now? Thanks in anticipation for ur kind response.
Thanks Waheed, for the positive feedback. Deeply appreciated! Read about Toda-Yamamoto procedure and deploy it.
Just a comment, you looked at the graph of the logarithm of the first difference of the series already in log, not just the difference of the logarithm. It actually doesn't change by much but it's not exactly the same.
Your comment is greatly appreciated, Erick! 🙏
@@CrunchEconometrix thank you for your response and your video!
This is great
Thanks for the positive feedback, Udo...deeply appreciated!
- First Diff. and Intercept
Null Hypothesis: D(LNPCE) has a unit root
Exogenous: Constant
Lag Length: 2 (Automatic - based on AIC, maxlag=8)
t-Statistic Prob.*
Augmented Dickey-Fuller test statistic -3.288906 0.0185
Test critical values: 1% level -3.510259
5% level -2.896346
10% level -2.585396
in your video you say this has unit root and reject the Null. but later you say this is acceptable. pls. explain this
Please understand this video. The level of series has a unit root. The 1st difference does not. I gave clear explanations in all cases. Kindly watch and take notes, thanks.
Thank you :)
You are welcome, Sualeha! 🙏 ❤️
Thank you so much for the amazing videos.
I just subscribed to your channel.
Could you please recommend a good and explanatory text book to read.
Thank you.
Thanks for your subscription, Angela. Deeply appreciated! My go-to textbooks are (1) Gujarati Basic Econometrics and (2) Woodridge Introductory Econometrics. You can never go wrong with those 2.
Thank you Dr. Please my data is a panel data and I have unbalanced please any video how to fix using eviews.
Hi Abdirahman, using unbalanced panel is ok unless there are TOO many missing observations then you might need to drop those variables and replace with closer proxies.
OK,many thanks for your help
You mentioned in other videos that “VAR must be specified in levels, hence VAR in differences would be misspecified”. Does this mean stationarity doesn’t matter for VAR? Thanks for the videos, very much appreciated from Canada.
Hi Bayo, I gave the references to the two schools of thought. I follow the first...thanks for the positive feedback. Appreciated.
Thank you so much your videos are really helpful. one should check stationarity after taking log of a series of without log? What is the benefit of taking log of a series
Hi Ruchi, read up on benefits of taking log of a variable. There are several resources online. Perform the ADF test on the form of the variable you intend using to run the analysis.
Hi! Thank you for the videos! I just want to understand one thing: I learned that when the series have a trend, sometimes I can turn it in a stationary one by calculating it´s first difference or logs. Therefore I though that testing the unit root in level, including trend and intercept should give the same result as testing the unit root in first difference including just the intercept (I mean that I excluded the trend because I´m using first difference). I´m right?
You can either include/exclude trend when testing for a unit root.
There's no reason to include the trend in the first difference unit root test
thanks for your responce
You are welcome, Sir.
please tell me that the idea behind test option with tend, intercept and both with trend and intercept !!!! is the variables all to be tested in intercept or in both with with software
Hi Abde, I often conduct with INTERCEPT and if still nonstationary, I include TREND.
Greetings Ngozi
thank you so much for this video. it helped. kindly, show how to run a unit root test in eviews, using 8 variables as a group or so.
kind regards,
Tjikari
It's the same procedure, Tjikari. Test each series individually.
...and thanks for the kind remarks on my videos. Humbly appreciated! Kindly tell others too by sharing my video links....gracias!
cool, thanks
Please, I ask if i want to run cointegration, must all variables be stationary in I(1) for the three models(Constant/ constant & trend/ None)? and also for the residuals model it must be stationary for the three models(3-2-1)?
Hi Zackaria, this video clearly explains how to conduct the stationary test using the Augmented Dickey-Fuller procedure. You may need to check other online resources if the information you require is not covered. Thanks.
Madam, Thank you very much for your prompt reply for my previous questions. I have a question on ADF test. My variables (dependent as well as independent) are showing unit root (i.e., non-stationary) in Level. However, they are showing as stationary in First difference as well as in the Second difference (for intercept and also intercept and trend) based on the p-value of 0.000. Then is my series is I(1) or I(2)? What should be the conclusion?
Hi Manoj, if you follow my explanations you can decipher the outcomes. I gave very clear explanations and interpretation of the ADF result. Thanks for watching and let me have the feedback.
Please how do we estimate a Linear Transfer Function in eviews?
Hi Obed, never heard of that technique. You may want to check out other online resources.
madam, many papers has used p-values in making decision of rejecting null hypothesis... here we compare ADF t-statistic and critical value. so my confusion is should we not consider p-values for decision. and thank u soo much for sharing your knowledge.
Hi Melina, yes you can use the pvalues too...and thanks for the positive feedback and kind words, deeply appreciated! Please may I know from where (location) you are reaching me?
@@CrunchEconometrix thank u so much for your prompt response, i am from Nepal :-). your tutorials are so helpful and i suggest every researcher i know to follow this. one more confusion i have , My GDP is stationary at level with trend and intercept and also significant at 1st difference with both intercept and trend and intercept. someone suggested me better to make decision of rejecting hypothesis on intercept rather than trend and intercept. is that so?
@@melinakharel4837 If stationary at level why take the 1st difference?
Great!
Thanks Leander!
thanks for all of your videos.... but is there any video on PP Unit root test, for structural breaks in time series data ?
Hi Abhijit, PP test is not suitable when there is a break in the model.
@@CrunchEconometrix perron test for structural breaks??
please what can i do if one of my raw time series data is non stationary even after taking its first difference.
Angela, you may have to change your estimation technique. Perhaps, use the Toda-Yamamoto technique. Check other online resources about the TY approach
@@CrunchEconometrix thank you so much for your response. So i changed my variables and they were all stationary at first difference I(1). Can i still use the ARDL model to estimate it?. Also, if i get an ECT of -1.3, considering it is more than 100% can i still say there is convergence? And finally, if one or two out of my four variables do not appear in the short run dynamics along with the ECM, what does this indicate and can i still interpret the other short run variables and use the results?
I have a question. Why didn't you use the "none" option in the ADF test for the differences of the variables?, and why should I have to use the max lenght of 8? My outcomes changes when I try with other maximum of lags
Hi Francisco, that Is because I want the CONSTANT to appear in the underlying equation. Use of lags is subject to several factors. I will refer you to watch my video on OPTIMAL LAGS.
Thanks all lot.
U're welcome Bunmi...may I know from where (location) you are reaching me?
@@CrunchEconometrix Ado Ekiti, Ekiti State
@@bunmiaugustina3323 Awesome Bunmi! 💕 😊 Kindly share my CZcams Channel link with your friends and academic community. I need many 🇳🇬 to be aware of my Channel. They'll learn some useful analytical skills and tips. Thanks for sharing! 💕 😊
Hello
I have a few questions. I am supposed to run a multiple linear regression. i'm an Eviews newbie.
- Is the unit root test supposed to be done on all variables?
- How do you proceed to the multiple linear regression after conducting the unit root tests?
Hi Anj, stationarity test is a MUST in time series analysis to avoid spurious results. My videos on VAR and ARDL are on multivariate analysis, kindly watch them....and kindly share my videos with your friends and academic. Thanks! 💕 😊
hello Dr. Sorry for asking, to plot the series, is it we always need to choose Log Difference, or we can use 'Difference' only?
Log difference is the growth rate of the series. Advisable to use the difference, although both plots look identical.
@@CrunchEconometrix all right, noted. thank you so much.
Hello MA, please how can I get eview software. I'm unable to get it
Hi Durojaye, EViews is not an open source software. You will have to buy it and install.
@@CrunchEconometrix Thank you ma. May I know which other software I can use for ADF unit root test and panel regression that is free for student to use in their research project.
You can use R. It's an open source (i.e. FREE) and robust for econometrics research.
What happens if the adf test rejects the null hypothesis with constant, and then rejects it with constant and trend?
It implies the series is stationary....and with a trend.
Hi mam!
Thankyou for the explanation, I would like to ask. Is the intercept and trend in ADF test is affecting my model choice in ARDL model?
Hi Emira, the ADF is to test for the presence of a unit root in a series. I'm not sure how that affects the ARDL model. Except that the outcome tells you whether using the ARDL approach is applicable or not.
@@CrunchEconometrix ohh i see,
Thank you mam
Hello Dr. Ngozi. Please do you have a video detailing the next step to take with the outcome of your unit root test? If yes, please could you forward me the link? i will very much appreciate it if you do. thanks
Hi Kenechukwu, my videos are sorted into 9 Playlists for easiness, please check through. It is not time-efficient sending links. Thanks.
Thank you for the videos. I have a question. How to determine the specification of the ADF test (intercept, intercept&trend, and none)?
And in my case, I have 3 variables. Two variables are stationary in 1st diff with specification none (no intercept&trend). And one variable is stationary in 1st diff with specification intercept. So can I conclude that all the three variables are stasionary at 1st diff?
Thanks in advance.
Yes, Azizah...we will conclude that all the variables are stationary after 1st difference.
Thank you so much for your reply. Really helpfull.
I find it very strange: If I use all options: "none", "intercept" and "trend & intercept" I get variables of the same order of integration.
Not strange at all. It happens sometimes depending on the variable used.
Hi, when I tested my data it came stationary in the first difference, but when make the estimation for the equation all of the variables are significant without converting the data to first difference, shall accept the model as its? or I must convert the data to first def. before applying the estimation? Thanks A lot.
Khaled, it is important to test for stationarity and the outcome directs subsequent procedures.
Pls how do i remedy a variable that is stationary at 2nd difference?
Kenechukwu, to the best of my knowledge such variables cannot be remedied. I'll suggest you drop it and replace it with a closer proxy.
@@CrunchEconometrix thanks for your response. Pls, what closer proxy can one use for HDI (human development index)?
Kenechukwu, as a researcher, you will have to do a search through the literature for this.
Hi Mam, hope you are okay, If variables are stationary in 1Dif. in PP test, and all stationary except one in DF test, can we accept PP alone, can we say that the data is stationary? thanks
Hi Khaled, yes you can. It shows that the stationary is evidenced from one of the tests.
Hi miss, I would to ask if is it compulsory to make the date become log? As like I already perform the test of ADF, PP, and Kpss in the first order.
Hi Tavia, taking log of a series is at the discretion of the researcher. Find out more about this online. Please may I know from where (location) you are reaching me?
Good day, I try to find the stationarity of my data using ADF test, and I got my t statistic value lower than all test critical values without differencing it. Does this mean my data is stationary? Just out of curiosity because I've seen quite a lot of the samples got higher t statistic value than test critical value in non-differenced level, and got lower t statistic value than test critical value in 1st or above level differenced data. Thanks in advance.
Hi Chan, this video is simple to understand and well-explained. I will encourage you to watch it again, jot down some points and align with your data. Best regards.
Hi ! i just discovered your channel thank you for this video :) i just want to ask you a question. there is this variable i'm working on, and when i do the ADF test with trend and intercept it shows that my serie is stationary and the p-value is 0.000 is it normal ? to be honest i started using eviews this week so i'm not familar with it yet.
Hi Sam, thanks for the positive feedback. I'll advice you to watch this video again and listen to my explanations. Read textbooks too. Understanding is very important. Please may I know from where (location) you are reaching me?
thanks for your explanation! by the way i'm a lil bit confused on how to conclude if a variable is stationary in 1st diff using the "none" specification, but it's stationary in level using "trend"? is the variable stationary in level or 1st difference?
Nadia, once my variable is stationary by "none", OR "trend" I'm good to proceed with my estimation. All I do is to indicate what I did in my manuscript...case closed.
hi very informative video, however i have one question. One of my variables has a unit root and when differenced at the first level, the data is stationary. Do i have to do the same for all my variables before i run my OLS, or can i proceed with my regression with only 1 of the variable differenced
Yes Hassan, perform URT for all the variables. Please may I know from where (location) you are reaching me?
@@CrunchEconometrix I have one variable at I(1) and the rest are I(0). Is basic OLS sufficient or would you recommend another model?
@@CrunchEconometrix Thanks for your reply. I am from London and my course leader had recommended that the students use your channel for econometrics guides.
@@hassanmahmood6584 No. Use ARDL.
@@hassanmahmood6584 I'm wowed and humbled by his recommendation. God bless him greatly. May I know the Institution?
Hi, what about panel data??
How do you mean?
First of all,Many thanks for your efforts,please i'm from algeria ,i would know how use the variables once you make a test of stationarity ,in your example lnpce is stationary at first difference with only constant if i run a regression i should use a new variable (lnpce-alpha d(lnpce)-constant) instead of lnpce,please i want more explanation ,
or if i have another variable stationary in first difference constant and trend i have to run regression and make modification for the original variable to use it in regression???
Many thanks
Hi Aziz, I'm sorry that I did not quite understand your query. However, how you use the variables depends on your estimation technique.
Dear Dr Ngozi, what criteria does one use to know whether to use intercept or trend and intercept in unit root test?
All models should have the intercept. Include the trend only if you observe such in the series or if the trend coefficient is statistically significant.
Hello Teacher...I have a question...is it possible to conduct unit root test,if sample size is 10?If not, what should I do?
Hi Antara, not at all. URT not applicable if time span < 15 years.
Hi Mam Again, thanks for your last answer to my question, I won ask. I am having 6 variables in my study, when I applied the unit root test, 5 of them are stationary on the first level, but the 6th one is stationary on 2nd level, Can I proceed within the regression model?
Hi Khaled, kindly watch my video on the Bounds Cointegration Test. Thanks.
Thank you so much for this tutorial. Could I ask you some question? I have annual data (2003-2017) because quarterly data is difficult for me to select in some countries. I want to choose a suitable las length of the variables (Export/Import, REER, RealGDP and Trade Openness). So, I found that lag length 3 is appropriate for REER variable. When running ADF test, REER is non-stationary with lag length 3 and it is stationary with lad length 1. Could you give me some advice about my situation, please? Thanks for your help!
Hi Helen, the reviewer will query the number of observations. You have 15 years, using 3 lags reduces the number to 12. To estimate any TS regression you must have AT LEAST 30 years observations to avoid "small sample" bias.
@@CrunchEconometrix Thanks for your help!
Thank u for the video. I wanna ask if I already have all variable is I(1) or in the first difference. What method can I choose?
ARDL or VAR depending on the objectives of your study.
Good day ma.
What do we do when the a particular variable is stationary at levels and first difference? Do we choose the one with the higher prob. value? Do we choose any one of the two? Do we choose the one with a higher ADF value?
Hi Siji, it is advisable to use the variable at its point of stationarity (either level-stationary or first difference-stationary) to avoid over-differencing. Though, some researchers argue in favour of such practices perhaps due to the need to have all variables at first-difference.
If the order of integration of variables where X is I(o), Y is I(1) & Z is I(2), what is the Appropriate Model for estimating the relationship among the variables.
Use Toda-Yamamoto technique.
God bless
You too, Wisdom...thanks!
Thank u very much, i would like to inform u that the link to the raleted topic in ur blog in the description section isn't working.
Ok, I'll check it out.
hello Mrs Ngozi, if my variables are only stationary at second difference i. e I(2), is there an estimation technique that suites that peculiar situation?
Hi Ola, that'll be the Toda Yamamoto approach which I've never used before. You may need to find out more about it online.
CrunchEconometrix thank you ma for replying. I'll check it out.
I wanna ask about the unit root test in second level difference, what is the best method are used? Are the method was use ARDL or VECM
Hi Henry, ARDL and VECM are not applicable. Use the Toda-Yamamoto technique.
@@CrunchEconometrix ok thank u
If the data is stationery at both level and first difference, can I used it as I(1) ? because I want to test Johenson test with other I(1) variables. Please answer teacher.
Hi Sithu, since you have a combination of level and 1st difference variables then ARDL is the most applicable model. I have several videos on the ARDL procedure, watch them and you will know what to do.
Hi dear! I love this tutorial too. I did the equation estimation and realized my r-squared was less than the Durbin-Watson stat which indicate my series is stationary. However, performing ADF unit root test at 'use level' indicate there is a unit root in the series. I'm confused now, please assist me. Thanks
Hi Bella, you are missing the point. Stationarity test is what validates if your series has a unit root or not. R2>DW only informs that you have performed a spurious regression by engaging nonstationary variables in a regression. Hope this clarifies. Thanks.
@@CrunchEconometrix I see. Thanks very much for the clarification.
Thank you so much for this tutorial. must all the series be transformed to their natural log before you conduct the stationary tests?
Not at all, Zama. Their raw forms can be used only that log-transformations smoothen the series.
Yes
@@CrunchEconometrix what determines the choice of maximum lag when conducting unit root test
besides ADF is it possible to use PP? i assume many paper use ADF and PP. but my vaiables shows different number what i mean is that let's assume there is variable x : even with first difference still non-stationary with ADF test but same variables with PP test from the stationary it is statble. in that case can i just use pp to all my variable instead of ADF? or is it possible each variables takes different test equation : variable x with trend variable y with non etc. i guess many ppl also will be confused whether each variables can change under their situation or do we have to unify all ?
thank you! with you help and video . I'v learned a lot nowadays . really appropriated
rina lee Very true, Rina you can use PP test in addition to ADF test... I do it often because just like you said it is very possible that variable Y may be nonstationary in ADF but stationary in PP, so it's good for researchers to report both results.
rina lee You are absolutely correct, Rina! That's the way to go... and make sure you state this clearly in your work so that any Reviewer will know what you did. Good job... keep me posted on the final results. Weldone!