UA-cam 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.
What do you say about researchers who think that lag number should be selected to satisfy the assumption of residuals being free of autocorrelation? I ask because, according to the below source, as implied, if the lag number suggested by any known information criteria generate VAR residuals with autocorrelation, the researcher may follow Hendy et. al. to use the lag number which frees the model of autocorrelation. For example, for a time sries instead of 1 or 2 lags, a researcher may use the lowest lag number which frees the model of autocorrelation. Please kindly clarify for me what your position is on that. Hendry, D. F. and Juselius, K. (2001), `Explaining Cointegration Analysis: Part II', Energy Journal 22(1), 75. Ji sie ike.
Good day Ma. Thanks for your teachings. They have been very helpful. Please my question is,can the bounds test for Cointegration be performed if you are not using the ARDL method?
Hello Prof. Thank you for your very useful educational Videos. Using Eviews 10 for analyzing ARDL model, there are five trend specifications (1-None, 2-Rest. Cons., 3-Constant, 4- Rest. Trend, and 5- Const. and Trend). Is there a method to decide which one is the best fit to the model in ARDL?
My journey in econometrics is becoming easier and enjoyable due to your tutorials; thanks again. Please may I know what steps need to be taken when the F-statistic is between the I(0) and I(1) bound?
Hi Bella, it's really heartwarming hearing such feedback. Thanks a bunch! Your query: if that's the case it means the outcome is inconclusive, hence you have 2 options: (1) estimate the short-run (ARDL) model or (2) change your regressors and begin estimation all-over again.
Hello Professor, I am from the UK, currently doing my thesis and I must say you are a life saver! Thank you so much for all the videos, they have been really helpful! I had a question, all of my variables are stationary at I(1) apart from one variable (inflation) which is stationary at I(0). Do I still have to use estimate bounds test ? When I did so, I found cointegration and I am not sure If I should use VECM or ECM ?
Hi Pooja, thanks for the encouraging feedback, Deeply appreciated. This video is well-explained on the steps required. You may also watch my video on "This is how to specify ARDL" on the steps to engaging the ARDL/Bounds/ECM. Thanks.
Thanks for great explanation but i have one doubt, if no long run relation exists according to the bound test result, then what are the short run coefficients
Very well explained but I have a doubt. Out of 5 variables, two are coming out to be 'not-cointegrated' and the rest 3 are coming out to be cointegrated. In this case, which test should I apply?
Gauri, it appears that you are yet to understand the ARDL-Bounds test. Not performed on variable but on the MODEL. Kindly watch the video on "This is how to specify the ARDL Model".
Thank you ma….I have a question, so I did unit root test for my variables, they were all non-stationary at level but at first difference, two of the variables became stationary, I’m confused on the regression method to use….maybe Ardl or Ols I would really appreciate your response ma
Dear Prof, i have four variables: A, B, C, D. Variable A is I(0), variable B, C, D is I(1). That's means based on your guide I should use ARDL and NOT the Johansen test since the order is different right? Then I run the ARDL equation and do the bound test. equation A, F-stats higher than Upper bound > means there is cointegration so I need to do ECM. Equation B, also show cointegration and needed to do ECM. Equation C show no cointegration then I just run the ARDL short run model. When I finally do the equation D, the F-stats results between the lower and upper bound. Based on your tutorial that's mean the result is inconclusive. What should I do with the equation D? Should I leave it and only do the other 3 equation? or is there any alternatives?
Hello Doctor, hope you are doing well! I have 2 questions: 1)Why the Short run coefficients are always small than the Long run ARDL? 2) I want to make forecasts using the ARDL model and I know I will be using the short run model but do I need to add the error correction term as well?
Thank you ma….I have another question, so I did unit root test for my variables, they were all non-stationary at level but at first difference, two of the variables became stationary, I’m confused on the regression method to use….maybe Ardl or Ols I would really appreciate your response ma
Thank you very much for the video. Its really helpful. I have a question. Some of my variables are I(2) other are I(1) and I(0). Can I still use this lmethod?
Good Videos indeed. I am working on time series data (aggregated data) with only 25 observations and I have learnt that for results to be reliable, at least I must work with 30 observations. Since it's not recommended to disaggregate time series data into quarterly or monthly frequency and I could not go back and get more data. How best can I deal with the situation and produce reliable results? I am working with an ARDL Model. Any help please.
Thank you ma for the video. Regarding the decision rule for the bounds test, the F-statistic in your example is clearly lower than all the critical values so what if the F-statistic is greater than some values in the I(0)?
if the test is inconclusive i.e the f-stat falls between the I(o) and I(1) values. what method do we use for estimation and what would be our conclusion
Thanks for the video. I have few questions. How levels equation is related to Granger causality? Significance level in levels equation means what? Do signs of levels equation is a method to verify theoretical insight? Thank you very much.
Thank you for these useful videos. I have a doubt. Presently am doing a time series project. Is it necessary to make the time series data normally distributed before doing Johansen Cointegration. I tried to normalise the data by using log transformation, min-max method, z-score method, square root as well as cube root method but Jarque Bera test probability is still 0.00000, it is not improving in any way. What should I do to normalise the data? Please help. Thank you
Thanks very much ma... I tried using this method but return near singular matrix... Because the dataset is a panel data.. 5 by 5....cross section is 5 and the number of years also 5 years... At this junction, what's the best model to apply for this panel data ma and what can I do to run this data sourced.... Investigating the effects of credit risk management on banks performance.. Sample size of 5 banks within 2016 to 2020.... Thanks in anticipation of your kind considertion and subsequent guidance on this research work ma...
Hi, That was wonderful presentation. pLease what VERSION of Eviews did you use for this estimation. Is it: Single user full license or university Edition or commercial volume license or academic volume license? I have standalone EViews 11 version but I could not perform the bound test because the long run/bound test is not on menu bar after clicking on coefficient diagnostics. Thank you.
respected, i have my model with mix cointegrationi(1) and i(0). i applied ARDL bound test but the result of lon run are positive but insignificant ,how can i interpret this long run ....tnx
Hello professor, may I ask should i log my data before doing ardl? As I have 1 variable set of data's in which after log it, the data is insufficient, so can I just log other variables and just left this insufficient data's variable without logging it?
Hi Yin, it is NOT compulsory that you log ALL variables. You can estimate a model using both logged and unlogged variables. You have the discretion to decide to best approach given the data you have.
I'm running an ardl model of n=30 and five variables . My Fstat comes out too big , more than 100, sometimes even 80. Is this normal can I go on and claim co integration ?
thanks for you videos, but in all your videos you went through all details without explaining each table in Eviews results refere to what (short run, long run...etc).... we need to know which tables we have to use in our research papers ...i'm so much confused about that
Mohamed, thanks for the feedback. I ALWAYS do my best to cover the basics with interpretations and I hope you will appreciate that. Also, know that you CANNOT get all you require from a video. Same way you need several econometrics textbooks to understand any topic/technique. Please support my videos with further readings. Keep watching, keep sharing...thanks!
Dear Prof. In Unit Root Test after choosing ADF there are 3 tests (Intercept, Trend, and None) for each unit root test (Level, 1st difference, or 2nd difference). how many of them should be significant to decide if a variable is stable at (Level, 1st difference, or 2nd difference).
Hello,Teacher.I want to ask you a question.Suppose some of the variables in an econometric model are not stationary at level.Even those are not stationary when those are transformed into logged values.But they are stationary at first difference. Which values should be used for co-integration test? Logged values of raw variables?or values after first difference? Should I use logged values of raw variables even if those are non-stationary?
Hello Ma'am, I have a similar question. I have two variables achieving stationarity at different levels I(0) and I(1) let's say they are called PF (independent)and MB (dependent).when performing the bounds cointegration tests in eviews, do I cointegrate d(MB) and PF? or is this not necessary? When I cointegrated MB and PF my results showed no cointegration but when I cointegrated d(MB) and PF there was cointegration
@@temiolanipekun6579 Bounds cointegration test is performed after the ARDL procedure using the same variables. I've always said that if you are using OLS algorithm, use d( ) but if using the ARDL algorithm use the variables as they are. Hope this clarifies. Thanks.
Hello teacher, your instructional video is very clear and useful. I have a question now. My f statistic is very high, as high as 32. Is this normal? Can I continue?
That is a really interesting stuff. What if I have a mix of variables that are I(0), I(1) and I(2). The similar co-integration technique may be used or there is some other test?
Hi Irfan, the Bounds cointegration test is no longer applicable with the presence of I(2) variables. You may need to deploy the Toda-Yamamoto procedure. Kindly check online for more information about this. Regards.
Hi mam..can you please clarify about short run interpretation in ARDL model as I have taken 3 lags and I am getting the coefficients of all 3 lag values..which one to mention in while writing for my research paper?? Thanks in advance
@@CrunchEconometrix mam will it be correct to interpret all coefficients of variables whose probability value is not significant..I have read various articles related to ARDL approach, they have mentioned only 1 coefficient of each variable. Please help me with this mam.
@@CrunchEconometrix thank you Madam. So, EC (levels equations) is different from ECM?? What do the coefficients in the LR mean? For example, if the F-statistcs show that there is cointegration, may I interpret the coefficients as the same I Do in the short run?
Hi teacher. Let me ask a short urgent question. When I do bounds test F value shows 1% significant. But its absolute t stats is below the lower bound. Anyway I assume it has cointegration. So I proceed error correction form. When I check the result, No long run coefficients are significant at all. Then I change the regression method, OLS by means of extracting residual. Under this method most short run coefficients become significant which do not in the former model and also found different result for ect and long run coefficients. Here, my question is (1) what should I do if absolute t stats value against the result of F value. (2) can I use ARDL in both OLS method and ARDL method alternately. (OR) which method is better. Thanks a lot.
Hi Sithu, I've responded to your several previous queries on different comment sections of my Channel. But I'll let this pass...too long a query. My advice: watch all my ARDL videos, jot down some keynotes and you'll know if u're on track or not. Thanks keep watching and tell your friends too.
Thank you for the videos. but, I am having some confusion regarding to determine the max lags for independent and dependent variables. I want to know how to select the max lags for independent and dependent variables. There are 34 observations in my data. Can you explain to me regarding this matter.. Thank you maam.
Sam, the literature on the "status" of the depvar in the ARDL is not quite clear. But, that the depvar is I(1) is in order. Nonetheless, several studies have used I(0) depvar.
Im new to cointegration concept. Can i use this bound test for pannel data too? If not then which cointegrate test i can conduct for pannel data (combination of I(0) and I(1)?
Madam, I got the F-statistic value of 4.0555973. At 1% level of significance, I(0) is 4.29 and I(1) is 5.61. Can I conclude that there is no cointegration (since F < I(0)).. However, this F-statistic value is higher than the critical values of I(0) at 10%, 5% and 2.5%. Is the conclusion of no cointegration alright? Please let me know.
Madam, Do we need to use the critical values of I(0) and I(1) at 5%, 1%, 10% as given by Eviews or we need to use the critical values as given by Pesaran and Narayan at 5%, 1% and 10%?
What if after unit root test you get outcome 3. do ARDL but find that the ARDL is non parameterized so you go on and do Wald test and end up with OLS which is specific parsimonious equation, which method do i use to conduct cointegration from here(parsimonious equation)? bound test or johasen? kindly reply.
@@CrunchEconometrix Thank you so much Professor for responding. I am from Delhi,India. Pursuing PhD (Economics) and your videos are really helpful to me. Regards
@@apica1234 Awesome!...and wish you the very best. Kindly share the link to my UA-cam Channel with your friends and academic community in India 🇮🇳 for awareness...thanks 😊.
Could someone please help me find and download the dataset used in example? The given gdrive link in the video description is not working .Im unable to access dataset here. Urgent help needed. pls.
Hi Kailash, I deactivated the link due to the unethical behaviors of people who attempted hacking my Google drive on 6 different occasions. Hence, I changed the access policy. Some datasets are free while some are available upon payment. Kindly check the link to find out which cruncheconometrix.com.ng/shop/
@@CrunchEconometrix Thanks a lot for the prompt reply M`am. Yes, I found the datafile listed here on your blog. Here cruncheconometrix.com.ng/product/crunch_dar-xlsx/. . and its not freely available for download. I could have really paid for this is if i was an employed professional and using this dataset for commercial purpose. But as of now im a student working on econometrics class assignment and am unable to pay anything for it. Is it possible for u to please kindly email share this file with me for educational purpose.? Any help will be very much appreciated. I promise to not post or share/ sell this data anywhere and use it for assignment only. Thank you.
Ghada, thank you for always encouraging me with your positive feedback. I'm grateful for the confidence in my videos. Kindly inform your colleagues and academic community on social media for awareness. They'll learn some useful tips and skills too. Once again, thanks! 💕 😊
Hi Glory, depending on your research objectives, you can estimate either a VAR or ARDL model. Kindly watch my videos on these. May I know from where (location) you are reaching me. Thanks.
@@CrunchEconometrix from university of nigeria, nsukka Enugu state.... What if you have three variables integrated in order one and the fourth variable order of two. What should i use? And what if at the end of my analysis my Durbin waston is 2,5
@@OluchiMbah Use Toda-Yamamoto procedure and know that there is more to diagnostics than the DW statistic. Read textbook on the DW and familiarize yourself with my videos on diagnostics. I'll appreciate if you can share the link to my UA-cam Channel with your friends and academic community in UNN and on the social media for awareness. Thanks 😊
UA-cam 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.
M`am, Could u please share a link to download the dataset used here. The given link is not working! Thx
What do you say about researchers who think that lag number should be selected to satisfy the assumption of residuals being free of autocorrelation? I ask because, according to the below source, as implied, if the lag number suggested by any known information criteria generate VAR residuals with autocorrelation, the researcher may follow Hendy et. al. to use the lag number which frees the model of autocorrelation. For example, for a time sries instead of 1 or 2 lags, a researcher may use the lowest lag number which frees the model of autocorrelation.
Please kindly clarify for me what your position is on that.
Hendry, D. F. and Juselius, K. (2001), `Explaining Cointegration Analysis: Part II', Energy Journal 22(1), 75.
Ji sie ike.
Check link here cruncheconometrix.com.ng/shop
Kindly watch my video on Optimal Lag Selection. Well explained. Thanks.
What a bombastic lecture.... Clear my 1000 confusions regarding ARDL Bounds...
Thanks, Abdullah for the encouraging feedback. Deeply appreciated!
@CrunchEconomitrics I am from SIMAD University Somalia, now I'm doing my thesis and I must say Thank U very much you made me easy
Thanks, Mohamed for your encouraging feedback. Deeply appreciated! 🥰🙏
Your classes have been very helpful. Thank you very much and well done.
Thanks for the encouraging feedback, MOG. Deeply appreciated 🙏🥰
Thanks very much. A very useful and practical tutorial. An excellent source of complementary learning
Glad it was helpful, Jose!
Good day Ma. Thanks for your teachings. They have been very helpful. Please my question is,can the bounds test for Cointegration be performed if you are not using the ARDL method?
The bounds test is related to ARDL modelling.
Hello Prof. Thank you for your very useful educational Videos. Using Eviews 10 for analyzing ARDL model, there are five trend specifications (1-None, 2-Rest. Cons., 3-Constant, 4- Rest. Trend, and 5- Const. and Trend).
Is there a method to decide which one is the best fit to the model in ARDL?
Hi Rebaz, there may be but I keep it simple and always use EViews default which is model 3.
My journey in econometrics is becoming easier and enjoyable due to your tutorials; thanks again. Please may I know what steps need to be taken when the F-statistic is between the I(0) and I(1) bound?
Hi Bella, it's really heartwarming hearing such feedback. Thanks a bunch! Your query: if that's the case it means the outcome is inconclusive, hence you have 2 options: (1) estimate the short-run (ARDL) model or (2) change your regressors and begin estimation all-over again.
Thanks for the insight. However, is it not necessary to transform the variables before the ARDL estimate?
Variable transformation is at the discretion of the researcher...advisable, though.
Hello Professor, I am from the UK, currently doing my thesis and I must say you are a life saver! Thank you so much for all the videos, they have been really helpful! I had a question, all of my variables are stationary at I(1) apart from one variable (inflation) which is stationary at I(0). Do I still have to use estimate bounds test ? When I did so, I found cointegration and I am not sure If I should use VECM or ECM ?
Hi Pooja, thanks for the encouraging feedback, Deeply appreciated. This video is well-explained on the steps required. You may also watch my video on "This is how to specify ARDL" on the steps to engaging the ARDL/Bounds/ECM. Thanks.
Thanks for great explanation but i have one doubt, if no long run relation exists according to the bound test result, then what are the short run coefficients
Simran, you will get them from the INITIAL ARDL estimation. Thanks
Very well explained but I have a doubt. Out of 5 variables, two are coming out to be 'not-cointegrated' and the rest 3 are coming out to be cointegrated. In this case, which test should I apply?
Gauri, it appears that you are yet to understand the ARDL-Bounds test. Not performed on variable but on the MODEL. Kindly watch the video on "This is how to specify the ARDL Model".
Thank you ma….I have a question, so I did unit root test for my variables, they were all non-stationary at level but at first difference, two of the variables became stationary, I’m confused on the regression method to use….maybe Ardl or Ols
I would really appreciate your response ma
Hi Rachael, I explained this in my ARDL videos. Use ARDL if all variables are I(1) or a mix of I(1) and I(0).
Hello Prof.
Please can you tell me what to do if all variables dependent and independent are I(1)?
Hi Shrishi, I explained this in the video titled "This is how to specify ARDL model ". Kindly watch it, thanks.
Dear Prof, i have four variables: A, B, C, D. Variable A is I(0), variable B, C, D is I(1). That's means based on your guide I should use ARDL and NOT the Johansen test since the order is different right? Then I run the ARDL equation and do the bound test. equation A, F-stats higher than Upper bound > means there is cointegration so I need to do ECM. Equation B, also show cointegration and needed to do ECM. Equation C show no cointegration then I just run the ARDL short run model. When I finally do the equation D, the F-stats results between the lower and upper bound. Based on your tutorial that's mean the result is inconclusive. What should I do with the equation D? Should I leave it and only do the other 3 equation? or is there any alternatives?
Estimate ARDL model for D. May I know from where (location) you are reaching me?
Can I only use the test if the dependent variable is I (1)? I read this information in a paper. Excellent video.
Yes, you can...and thanks for the positive feedback. Appreciated!
Hello Professor, this video is very helpful. But In my Eviews 10, I don't have the option of Long run form and bound test in coefficient diagnostics
Hi Swathi, both menus are standalone. Not under DIAGNOSTICS. What version of EViews are you using?
@@CrunchEconometrix Eviews 10
Swathi, I use EViews10. Watch my views and follow the steps.
Hello Doctor, hope you are doing well!
I have 2 questions: 1)Why the Short run coefficients are always small than the Long run ARDL?
2) I want to make forecasts using the ARDL model and I know I will be using the short run model but do I need to add the error correction term as well?
Hi Charifa, (1) I can't say why and (2) I have not used the ARDL approach to forecast. You may want to check out other online resources.
Good evening professor, please is the bounds test for each hypothesis?
Hi Rachael, the Bounds test is performed for each model.
Thank you ma….I have another question, so I did unit root test for my variables, they were all non-stationary at level but at first difference, two of the variables became stationary, I’m confused on the regression method to use….maybe Ardl or Ols
I would really appreciate your response ma
Thank you very much for the video. Its really helpful. I have a question. Some of my variables are I(2) other are I(1) and I(0). Can I still use this lmethod?
No you can't, Kristian. I mentioned this in the video.
Good Videos indeed.
I am working on time series data (aggregated data) with only 25 observations and I have learnt that for results to be reliable, at least I must work with 30 observations. Since it's not recommended to disaggregate time series data into quarterly or monthly frequency and I could not go back and get more data.
How best can I deal with the situation and produce reliable results?
I am working with an ARDL Model.
Any help please.
Hi Gabriel, data structuring is permissible. You may need to look up re-structuring yearly to quarterly or monthly data on how best to proceed.
Thank you ma for the video. Regarding the decision rule for the bounds test, the F-statistic in your example is clearly lower than all the critical values so what if the F-statistic is greater than some values in the I(0)?
Hi Sharon, the F-stat should be greater than the I(1) critical values to reject the null hypothesis of no cointegration as explained in the clip.
Ma'am!
F-Statistics value is 3.837083 do i reject null hypothesis?
Superb Video
Shabbar, the decision criteria is clear. Pls watch the video again.
if the test is inconclusive i.e the f-stat falls between the I(o) and I(1) values. what method do we use for estimation and what would be our conclusion
Your conclusion is NO cointegration.
Thanks for the video. I have few questions. How levels equation is related to Granger causality? Significance level in levels equation means what?
Do signs of levels equation is a method to verify theoretical insight? Thank you very much.
Hi Nura, thanks for the positive feedback but your queries are either inconsistent or unrelated to this clip. Don't understand what you mean.
Thank you for these useful videos. I have a doubt. Presently am doing a time series project. Is it necessary to make the time series data normally distributed before doing Johansen Cointegration. I tried to normalise the data by using log transformation, min-max method, z-score method, square root as well as cube root method but Jarque Bera test probability is still 0.00000, it is not improving in any way. What should I do to normalise the data? Please help. Thank you
Hi Aritra, transforming a variable may not result to normality.
Thanks very much ma... I tried using this method but return near singular matrix... Because the dataset is a panel data.. 5 by 5....cross section is 5 and the number of years also 5 years... At this junction, what's the best model to apply for this panel data ma and what can I do to run this data sourced.... Investigating the effects of credit risk management on banks performance.. Sample size of 5 banks within 2016 to 2020.... Thanks in anticipation of your kind considertion and subsequent guidance on this research work ma...
Simon, how can you post a panel data question on a TIME SERIES thread? I have several videos on panel data analysis. Kindly watch them. Thanks.
Hi,
That was wonderful presentation.
pLease what VERSION of Eviews did you use for this estimation.
Is it: Single user full license or university Edition or commercial volume license or academic volume license?
I have standalone EViews 11 version but I could not perform the bound test because the long run/bound test is not on menu bar after clicking on coefficient diagnostics.
Thank you.
Hi Chelsea, unfortunately I can't say which exact version.
Hey, which method of cointegration do I have to use if I'm using 4 variables and 3 were stationery at second difference, one was stationery in levels
Sharon, for 1(2) series, you can deploy the Toda-Yamamoto technique. You can check out other online resources for more information.
Excellent one. Very helpful
Glad to hear that, Abdul...thanks!
Honestly, you did a great job here
Thanks for the encouraging feedback, Oyegoke...deeply appreciated!
That was very helpful to me, Thank you.
Thanks for the encouraging feedback, Sir🙏
respected, i have my model with mix cointegrationi(1) and i(0). i applied ARDL bound test but the result of lon run are positive but insignificant ,how can i interpret this long run ....tnx
Hi Abu, insignificant coefficient implies NO impact on the depvar.
Godbless you prof!!! ❤❤❤❤
U're welcome, Sir! 🥰 🙏
Hello please i have a question. What should we do if the bound test is inconclusive ?
Hi de Safa, you can either estimate the short-run unrestricted ARDL model or change some control variables and re-estimate.
Dear Prof. Regarding ECM value, how many scenarios are possible. For Example is it possible to be greater than /1/? How to interpret them?
Hi Rebaz, assuming the ECT is -1.02, it is the same interpretation except that adjustment to long-run equilibrium is faster at 102%.
Hello professor, may I ask should i log my data before doing ardl? As I have 1 variable set of data's in which after log it, the data is insufficient, so can I just log other variables and just left this insufficient data's variable without logging it?
Hi Yin, it is NOT compulsory that you log ALL variables. You can estimate a model using both logged and unlogged variables. You have the discretion to decide to best approach given the data you have.
I'm running an ardl model of n=30 and five variables . My Fstat comes out too big , more than 100, sometimes even 80. Is this normal can I go on and claim co integration ?
The higher the value of F-stat, the better your result. I suggest that you read more about the essence of the F-stat.
thanks for you videos, but in all your videos you went through all details without explaining each table in Eviews results refere to what (short run, long run...etc).... we need to know which tables we have to use in our research papers ...i'm so much confused about that
Mohamed, thanks for the feedback. I ALWAYS do my best to cover the basics with interpretations and I hope you will appreciate that. Also, know that you CANNOT get all you require from a video. Same way you need several econometrics textbooks to understand any topic/technique. Please support my videos with further readings. Keep watching, keep sharing...thanks!
Dear Prof.
In Unit Root Test after choosing ADF there are 3 tests (Intercept, Trend, and None) for each unit root test (Level, 1st difference, or 2nd difference).
how many of them should be significant to decide if a variable is stable at (Level, 1st difference, or 2nd difference).
Hi Rebaz, kindly watch my video on ADF. It is well-explained. Keep it simple and follow my procedure. Thanks.
Hello,Teacher.I want to ask you a question.Suppose some of the variables in an econometric model are not stationary at level.Even those are not stationary when those are transformed into logged values.But they are stationary at first difference.
Which values should be used for co-integration test?
Logged values of raw variables?or values after first difference?
Should I use logged values of raw variables even if those are non-stationary?
Your queries are confusing. Please clarify.
Hello Ma'am, I have a similar question. I have two variables achieving stationarity at different levels I(0) and I(1) let's say they are called PF (independent)and MB (dependent).when performing the bounds cointegration tests in eviews, do I cointegrate d(MB) and PF? or is this not necessary? When I cointegrated MB and PF my results showed no cointegration but when I cointegrated d(MB) and PF there was cointegration
@@temiolanipekun6579 Bounds cointegration test is performed after the ARDL procedure using the same variables. I've always said that if you are using OLS algorithm, use d( ) but if using the ARDL algorithm use the variables as they are. Hope this clarifies. Thanks.
Hello teacher, your instructional video is very clear and useful.
I have a question now. My f statistic is very high, as high as 32. Is this normal?
Can I continue?
Hi Wl, the higher the F-stat the stronger the evidence to reject the null hypothesis.
@@CrunchEconometrix Thank you professor.Your teaching is wonderful.I will keep watching you.
...and please help to share too! :)
thanks a lot prof
You are welcome, Sir 🥰🙏
That is a really interesting stuff. What if I have a mix of variables that are I(0), I(1) and I(2). The similar co-integration technique may be used or there is some other test?
Hi Irfan, the Bounds cointegration test is no longer applicable with the presence of I(2) variables. You may need to deploy the Toda-Yamamoto procedure. Kindly check online for more information about this. Regards.
Hi mam..can you please clarify about short run interpretation in ARDL model as I have taken 3 lags and I am getting the coefficients of all 3 lag values..which one to mention in while writing for my research paper?? Thanks in advance
Hi Zoya, you interpret all. Watch my other ARDL videos for interpretations OR get any ARDL-based article and adapt their interpretations.
@@CrunchEconometrix mam will it be correct to interpret all coefficients of variables whose probability value is not significant..I have read various articles related to ARDL approach, they have mentioned only 1 coefficient of each variable. Please help me with this mam.
@@CrunchEconometrix please reply mam as its urgent...
PLEASE HELP-ME. What does the Error Correction (LEVELS EQUATION) mean? Is this the long run? I really need some help.
Yes Rafael, that is the LR result.
@@CrunchEconometrix thank you Madam. So, EC (levels equations) is different from ECM?? What do the coefficients in the LR mean? For example, if the F-statistcs show that there is cointegration, may I interpret the coefficients as the same I Do in the short run?
Rafael, get an ARDL-based paper and read Section 4 on results interpretation. You can get my papers listed in the COMMUNITY TAB. Thanks.
Hi teacher. Let me ask a short urgent question. When I do bounds test F value shows 1% significant. But its absolute t stats is below the lower bound. Anyway I assume it has cointegration. So I proceed error correction form. When I check the result, No long run coefficients are significant at all. Then I change the regression method, OLS by means of extracting residual. Under this method most short run coefficients become significant which do not in the former model and also found different result for ect and long run coefficients. Here, my question is (1) what should I do if absolute t stats value against the result of F value. (2) can I use ARDL in both OLS method and ARDL method alternately. (OR) which method is better. Thanks a lot.
Hi Sithu, I've responded to your several previous queries on different comment sections of my Channel. But I'll let this pass...too long a query. My advice: watch all my ARDL videos, jot down some keynotes and you'll know if u're on track or not. Thanks keep watching and tell your friends too.
Thank you so much
You are welcome, Rvn! 🙏 ❤️
I have 62 observations...should I take max lag 2? Or should I go with optimal lag selection with AIC as you have suggested in one of yours video
Becoz that 2 lag for my model with the help of AIC
Hi Ruchi, either way is good.
Sure, go ahead and use the 2 lags.
Take my humble greetings professor 💜 can you explain in a short note that when to use restricted constant and when to use unrestricted constant?
Hi Tabitha, there's a whole lot of explanation in there. My suggestion is that you get any econometrics textbook for in-depth understanding. Thanks
@@CrunchEconometrix Actually i couldn’t found such appropriate books....can you kindly suggest me one? 😓
Tabitha, if you do a Google query, you will get linked resources.
Thank you for the videos. but, I am having some confusion regarding to determine the max lags for independent and dependent variables. I want to know how to select the max lags for independent and dependent variables. There are 34 observations in my data. Can you explain to me regarding this matter.. Thank you maam.
Hi Ain, watch my video on "Optimal lag selection". Well explained. Thanks.
Hello. I have read somewhere that ARDL bound test can only be used if the dependent variable is non stationary at level. Is this true?
Sam, the literature on the "status" of the depvar in the ARDL is not quite clear. But, that the depvar is I(1) is in order. Nonetheless, several studies have used I(0) depvar.
Thanks
You are welcome, Abdulrazak.
Im new to cointegration concept. Can i use this bound test for pannel data too? If not then which cointegrate test i can conduct for pannel data (combination of I(0) and I(1)?
Watch my videos on panel ARDL.
@@CrunchEconometrix thank you.
Madam, I got the F-statistic value of 4.0555973. At 1% level of significance, I(0) is 4.29 and I(1) is 5.61. Can I conclude that there is no cointegration (since F < I(0)).. However, this F-statistic value is higher than the critical values of I(0) at 10%, 5% and 2.5%. Is the conclusion of no cointegration alright? Please let me know.
The rule for determining cointegration is quite clear. Apply it, and you have your answer.
Madam, Do we need to use the critical values of I(0) and I(1) at 5%, 1%, 10% as given by Eviews or we need to use the critical values as given by Pesaran and Narayan at 5%, 1% and 10%?
Hi Manoj, the results from both tables are not significantly different. Some studies use P&N if sample size is small, otherwise use either of them.
in this video i became know how to run the model but what about panel data co integration? :) same way? or different from this ?
rina lee Estimations for panel data are different. I'm already working on them. I'll roll out very soon... watch out! 😄
rina lee I'll do my best, girl. That's a promise! Which country are you based?
What if after unit root test you get outcome 3. do ARDL but find that the ARDL is non parameterized so you go on and do Wald test and end up with OLS which is specific parsimonious equation, which method do i use to conduct cointegration from here(parsimonious equation)? bound test or johasen? kindly reply.
I don't understand what you mean by "ARDL is non-parameterised".
how to proceed if 4 out of 5 equations show cointegration but 1 is inconclusive? how to specify the one which is inconclusive and estimate in eviews?
Hi Syed, that is simple. Follow my procedure and estimate 4 EC equations and 1 ARDL equation. May I know from where (location) you are reaching me?
If 3 individual series are stationary at level 2, level 2 and level 1... should we use bounds cointegration test ?
No Apica, Bounds test is not applicable. Use the Toda-Yamamoto technique. Read more about it. May I know from where (location) you are reaching me?
@@CrunchEconometrix Thank you so much Professor for responding. I am from Delhi,India. Pursuing PhD (Economics) and your videos are really helpful to me.
Regards
@@apica1234 Awesome!...and wish you the very best. Kindly share the link to my UA-cam Channel with your friends and academic community in India 🇮🇳 for awareness...thanks 😊.
Could someone please help me find and download the dataset used in example? The given gdrive link in the video description is not working .Im unable to access dataset here. Urgent help needed. pls.
Hi Kailash, I deactivated the link due to the unethical behaviors of people who attempted hacking my Google drive on 6 different occasions. Hence, I changed the access policy. Some datasets are free while some are available upon payment. Kindly check the link to find out which cruncheconometrix.com.ng/shop/
@@CrunchEconometrix Thanks a lot for the prompt reply M`am. Yes, I found the datafile listed here on your blog. Here cruncheconometrix.com.ng/product/crunch_dar-xlsx/. . and its not freely available for download. I could have really paid for this is if i was an employed professional and using this dataset for commercial purpose. But as of now im a student working on econometrics class assignment and am unable to pay anything for it. Is it possible for u to please kindly email share this file with me for educational purpose.? Any help will be very much appreciated. I promise to not post or share/ sell this data anywhere and use it for assignment only. Thank you.
More than excellent ..thank u professora this way is amazing now i absorb better
Ghada, thank you for always encouraging me with your positive feedback. I'm grateful for the confidence in my videos. Kindly inform your colleagues and academic community on social media for awareness. They'll learn some useful tips and skills too. Once again, thanks! 💕 😊
Please we want NARDL estimation
I'll do my best, Maaroufi...thanks!
Ma what do i do when all my values are integrated in first order?
Hi Glory, depending on your research objectives, you can estimate either a VAR or ARDL model. Kindly watch my videos on these. May I know from where (location) you are reaching me. Thanks.
@@CrunchEconometrix from university of nigeria, nsukka
Enugu state....
What if you have three variables integrated in order one and the fourth variable order of two. What should i use?
And what if at the end of my analysis my Durbin waston is 2,5
@@OluchiMbah Use Toda-Yamamoto procedure and know that there is more to diagnostics than the DW statistic. Read textbook on the DW and familiarize yourself with my videos on diagnostics. I'll appreciate if you can share the link to my UA-cam Channel with your friends and academic community in UNN and on the social media for awareness. Thanks 😊