Forum

Please feel free to post comments using the “Post your Comment” facility below to provide questions and answers relating to the subject of econometrics, research and data analysis. Since this is an interactive forum, I will not be able to answer all your questions myself, but other contributors can help out. I will “casually moderate” this forum to avoid spamming activities and inappropriate contents. We are here to learn, interact and have fun….so, let’s crunch this course!

Follow up with updates from CrunchEconometrix:

Website: http://cruncheconometrix.com.ng

 Blog: https://cruncheconometrix.blogspot.com.ng/

 Facebook: https://www.facebook.com/CrunchEconometrix

 YouTube Custom URL: https://www.youtube.com/c/CrunchEconometrix

 Stata Videos Playlist: https://www.youtube.com/watch?v=sTpeY31zcZs&list=PL92YnqQQ1gbjyoGWR2VUemNPU93yivXZx

 EViews Videos Playlist: https://www.youtube.com/watch?v=znObTs4aJA0&list=PL92YnqQQ1gbghRSJURtz08AZdImbge4h-

NOTE: Always click FORUM”,  BLOGPOST and YOUTUBE VIDEOS“ for updates on lecture and video tutorials.

104 Replies to “Forum”

    1. The application of statistical and mathematical methods to the analysis of economics data with a purpose of giving empirical contents to economic/finance theories and verifying them or refuting them.

  1. Woww!!
    This is a lovely platform. I trust what you can do. I believe this platform will also work together with Eduresearch.ng to improve the research skills of students and researchers in different field of study.

    1. I believe so too. We are here to teach the rudiments of econometrics (see BLOGPOST), encourage readers to engage in data analysis and also show the practical applications of using Stata analytical software with exciting videos….watch out for our YouTube videos!

    1. Yes, you are right that is the objective of setting up this site to demystify data analysis and the use of econometrics. We will appreciate if you can refer students to this blog so that they can follow the simplified tutorials, post their questions and other challenges relating to econometrics….and also attempt the assignments for better understanding of the applications of econometrics.

  2. This blog is long overdue. Crunch econometrix is set to demystify the monster in research called data analysis and interpretation. It’s a welcome development.
    Kudos to the Tutor.

    1. Thank you very much for the words of encouragement. You are encouraged to refer students to this site to enable them follow the simplified lecture tutorials on the blog. Simply click on BLOGPOST

  3. @ Engee thanks for this given back spirit in you, highly inspiring!
    Please how does the lecture follow? Or each independent of other? Because Heteroscedasticity came on Jan 8., follow by Multicollinearity, then Tell me, what is econometrics? Just for some of us starters in econometric worldview to be well grounded as we get along your laudable philanthropic teaching.

  4. Hi Gbenga, hope you’re doing ok. The lectures do not follow any particular order, they are independent of one another. But where lectures are inter-related, I will then prepare such tutorials in their almost particular order. For instance, I just posted on the one-way ANOVA (analysis of variance) procedure….and in the next couple of days, the two-way ANOVA procedure will be posted. This is because, it is important to understand the one-way ANOVA before gravitating to the two-way, three-way and so on.

    Also, readers often suggest some topics like two out the five posts are topics suggested by them (that’s http://cruncheconometrix.blogspot.com.ng/2018/01/heteroscedasticity.html and http://cruncheconometrix.blogspot.com.ng/2018/01/multicollinearity.html). So, you can always suggest any topic that require some simplified explanations.

    If I may ask: are you acquainted with any analytical software? Even I though I teach using the Stata application, I post datasets in excel files too so that readers can download and feed into any application of their choice to practice along as I teach. Which one are you using? Please let me know, thanks!

    1. Thanks so much @ Engee!
      Actually reason why I asked is because am just trying to learn econometrics and how to go about modelling. I’m a research student in psychology and I do analysis, and of recent some Agric Econs student seek my guide in econometric analysis which I don’t too know and I decide to do self-learning, I so much appreciate your passion and wish to grab sound knowledge in econometric from your expertise.
      I use SPSS, trying to learn “r” and will appreciate knowing Stata. Regards!

      1. Good to know that you are on track. The basics of econometrics are the same regardless of the analytical software. I’ll definitely help you out with using Stata. I’ll soon be doing some videos on the rudiments of using Stata. They are quite easy to grasp… just stay with me on this platform and I will do my best to teach the little I know to as many who are willing to learn a thing or two. Take care…

    1. Yes you can. But remember that cointegration implies that the series are related and can be combined in a linear fashion. So you can estimate a VECM and Granger causality. But if there is no cointegration, It implies that there is no long-run relationship so only the short-run model will be estimated. I hope this is helpful.
      Good luck with your research!

    1. Hi Liz,
      There’s the Human Capital Theory of Becker and Tomes (1979, 1986). It states that the decision of parents to invest in their children’s human capital development determines the extent of persistence of relative incomes across generations. The simple interpretation is that education reduces income inequality gap. As more people become educated, they are able to get better jobs and better incomes….read more on this online.

  5. Kindly upload a complete video lecture on Gregory and Hansen Cointegration Test with Structural Breaks using Eviews. It will be of a great help.

  6. Good day, please can you run a time series analysis with some missing observations in a particular data set? for example, if i have a time series data on exchange rate for Ghana and in some of the years i do not have data available for exchange rate, can i go ahead and run my estimation with what i have or i will have to fill the empty spaces before i do my estimation? thank you.

  7. Hi Nii, sincere apologies for the late response. Results from time series estimates are biased if the number of observations falls below 30 years (small sample bias). Unless, you are certain that after accounting for the missing data, obs>30 then you can go ahead, Otherwise, you need to drop the exchrate variable and find another proxy.

  8. Thanks for this insightful peace. Am not happy that I didn’t know of you or the site for long, but nevertheless am happy to find this site now. Thank you once again for the knowledge you share with us.

  9. Good afternoon admin… Thanks for your videos it has help me a lot. I am running an ARDL model and my stata can’t recognised the command (ardl) please can you help me

    1. …thanks Richard for the kind remarks. Humbly appreciated and kindly tell others too. Kindly post the error message you are getting on the comment section of the YouTube video that will aid in assisting you appropriately. Thanks.

  10. After reading your post, you have a great website with interesting content. But I think you can improve your current google ranks by using SEO website traffic net. My friend uses it and it works great. Just google it, it’s very nice tool to bring you a lot of new readers on a daily basis. Keep up the quality work!

  11. After reading your post, you have a great website with interesting content. But I think you can improve your current google ranks by using this service. My friend uses it and it works great. Just check to https://bit.ly/2v7j3id, it’s very nice tool to bring you a lot of new readers on a daily basis. Keep up the quality work!

  12. Hi Ngozi
    Thank you for your tutorials, they really help.

    Please help, how do I run a Unit root test using ADF for 8 variables.

    thanks
    Tji

    1. Hi Tji,
      Simply run the ADF test for each variable, You can watch my tutorials on how to do that. I have videos in Stata and EViews. Thanks for watching my videos, deeply appreciated! Please share my links with your colleagues.

  13. Hello everybody, I am running a VAR-model and closely follow the youtube tutorials in Stata on how to proceed. All in all, I have 6 variables and monthly data.
    The ADF-test shows that my variables are stationary after first differencing. The optimal lag length differs but it does not matter for my problem:
    When I try to use the autocorrelation test: “varlmar, mlag(#)”, the following message pops up:
    “the exogenous variables may not be collinear with the dependent variables, or their lags”
    Does anyone have an idea on how to interpret that or what I can do so that the message would not show up anymore?

    1. Hi Jan,
      6 variables are likely too many for a VAR model. Reduce lags to 1 and re-test for autocorrelation. Thanks for watching my videos, deeply appreciated!

  14. assuming that we use a dummy variable as one of the explanatory variables, for instance, say financial crisis and that a country experienced a financial crisis in 2008 and 2009. how can we use an ARDL model to show how the effect of the financial crisis in those 2 periods on the dependent variable say economic growth? what other ways can be used to show the effects of the financial crisis in those 2 years?

  15. assuming that we use a dummy variable as one of the explanatory variables, for instance, say financial crisis and that a country experienced a financial crisis in 2008 and 2009. how can we use an ARDL model to show how the financial crisis in those 2 periods affected the dependent variable say economic growth? what other ways can be used to show the effects of the financial crisis in those 2 years?

  16. Dr. This site has just renewed my interest in Econometrics and data analysis. I have tried to learn through people but fruitless. One major feature about them is that there are always not available to teach. My initial assignment would be forwarded for your input.
    Thank you maa.

  17. hey Adeleye hope to find you well.
    i am a phd Scholar from Pakistan. i am working with a dynamic panel with two data sets (balanced(N=29, T=30) and unbalanced(N=35, T=30).

    My model has one endogenuous variable as well, i was applying system GMM to control the issue of endogeniety because data on good external instruments is missing in my case. but due to large T my instrument set approaches 50-60 even with collapse option. You have suggested in one of your videos to use PMG, MG and DFE in large panels.
    My question is whether these models resolve the issue of endogeniety? or you can please suggest me some other technique for my data which not only gives consistent estimates but also resolves the issue of endogeniety in absence of good external instruments.

    waiting for your reply

  18. hi dear! I am Demisse from Ethiopia and im msc student in addis ababa university. this year im doing my thesis on the title energy-growth-environmental pollution taking selected east african countries. their selection is based on data availability. when i see the nature of my data, the correlation between regrassors found to be very high. even when i see correlation between combustible renewables and waste and fossil fuel energy, it reaches 99% and also i used to estimate my model in panel ARDL setting. one thing i would like to ask you is: is multicollinearirty a problem in panel ARDL model? is high correlation neccessairly infers multicolinearity?. thank you in advance.

    1. Hi Demisse,
      Multicollinearity is a problem in all regression analysis….high correlation between variables implies COLLINEARITY while high correlation among variables imply MULTICOLLINEARITY. May I know from where you are reaching me?

  19. Hi mam, I would like to ask is okay to add lagged independent variable in fixed/random effect regression.
    My equation is :
    PBR_it=α_0++α_1DEBT_(it-1)+β_1GAP_it+β_2WTI_it+β_3CC_it+ϵ_it

    T=16, N=7
    Is there any literature/book/research that confirm this type of equation?
    I’m looking forward for your advice, thank you

  20. Hello, thanks a lot for these videos they are very instructive. I however have a question about the model specification you show at 1.18 minutes of the video “how to estimate two step GMM”. There you include lnmob in both the iv()and the gmm() parts, however you do not do that in the dofile after, you only include that variable in either iv() or gmm(). Can you please tell me if this is a typo or is it correct? Also, if I already have much more groups than instruments in my data do I still need to include the collapse option? thanks in advance!

    1. Hi Justine, that will most likely be an error. So, follow what I did in the video. Since N>instruments, no need to use the collapse option.

  21. Hello, i followed keenly the tutorials on Panel Ardl. however, i would like to know how are the lags orders are included in the estimated model when doing the estimation. I have not seen anything that indicates that the lags orders are taken into consideration in estimating the model. if they are not incorporated in the model, then why do we have to do the lag selection criteria so as to select the most common ones in our model?
    Thanks

    1. Hi Bismark, the panel ARDL model syntax is user-written that is why you rarely come across video tutorials on the procedure. Experience also shows that the model breaks down when higher order lags are used. May I know from where (location) you are reaching me?

  22. Hello,

    How do I asses the significanc of the variables (which appear on the very first part of VECM output in Eviews) in the Vector Error Correction model. In particular, which distribution shall I use to get the t-critical values and which tailed test shall I use? one sided or two sided? Thank you.

    1. Hi Martina,
      That first part relates to the long-run equation. The t-stats are directly below the coefficients. The distribution is always a 2-tailed test.

  23. Hello,
    Thank you for your video on carrying out ARDL using OLS approach. I would like to use this method to carry out my thesis on time series analysis. I have two questions:
    1. Do you have specific references I can use for the equations you used in the short, long-run model and error correction model specifications? I would like to include the references in my thesis
    2. Why are the lagged regressors only included i.e why are the contemporaneous regressors not included in the models?

  24. Hello
    I need a detail help of your video for Propensity score matching. Because your all videos are very easy to understand and learn. Please help me

  25. Hello all,
    It is my first time on this website. I really appreciate because videos dofile and datasets allowed me to well understand many things in applied Econometrics.
    God bless you Doctor, just continue to help like that.
    Sorry for because my english is not fluent, I am french speaking.
    Bertrand

  26. Dear Ngozi,
    hope you’re doing ok.
    I don’t have words to thank you for your videos in relation to Econometrics. Thank you for the great job and continue to do the same.
    I am doing a thesis research on the interconnectedness of Electricity Consumption and Economic Growth in Ethiopia to test the dynamic causality relation between Economic Growth/GDP and Electricity Consumption including other variables like FDI, Government Expenditure and Net Export. I am using a Multivariate Co-integration Model jointly with the Multivariate Granger Causality Test to avoid the problem of omitted variable bias with the help of STATA. The data will be based on 1988-2017, 30 years data.
    I have followed your videos and other references to construct my methodology as follows:
    Step 1. Specify the Model
    Step 2. Time setting of the data in the application set the data by using “tsset” syntax to perform the time series analysis, tsset year
    Do I need to divide all data by 1000 to make the data in Billions? Is it a requirement to work in billions?
    Do I need to generate log variables and why, I didn’t get it in your videos, you just showed a changed data. What are the things I need to do on the data before doing the analysis?
    Step 3. Perform Stationary Test: series must be of order one (I(1)) and not of order two (I(2)) by using Augmented Dickey Fuller Test (ADF), H0= Model is non-stationary,
    I tried to do this based on your video but I couldn’t get stationarity in the second step what shall I do? Some pieces of literature explain that ADF test has a disadvantage for small data.
    Step 4. Determine Optimal Lag Length (k) for the model
    Step 5. Estimate VAR or VECM model
    Step 6. Perform diagnostic tests
    Is there a reference that I can use for the analysis like for example a STATA do file? The reference can be in areas of electricity or Energy Consumption and Economic Growth. I just want to understand the steps and some commands.
    Thanks, dear.

    1. Hi Ephrem,
      Sorry for the late response. Kindly reduce this. It is too long. Then post on the appropriate YouTube video and I will respond from there.
      Thanks.

  27. Dear Ngozi,
    hope you’re doing ok.
    I don’t have words to thank you for your videos in relation to Econometrics. Thank you for the great job and continue to do the same.
    I am doing thesis research on the interconnectedness of Electricity Consumption and Economic Growth in Ethiopia to test the dynamic causality relation between Economic Growth/GDP and Electricity Consumption including other variables like FDI, Government Expenditure and Net Export. I am using a Multivariate Co-integration Model jointly with the Multivariate Granger Causality Test to avoid the problem of omitted variable bias with the help of STATA. The data will be based on 1988-2017, 30 years data.
    I have followed your videos and other references to construct my methodology as follows:
    Step 1. Specify the Model
    Step 2. Time setting of the data in the application set the data by using “tsset” syntax to perform the time series analysis, tsset year
    Do I need to divide all data by 1000 to make the data in Billions? Is it a requirement to work in billions?
    Do I need to generate log variables and why, I didn’t get it in your videos, you just showed a changed data. What are the things I need to do on the data before doing the analysis?
    Step 3. Perform Stationary Test: series must be of order one (I(1)) and not of order two (I(2)) by using Augmented Dickey Fuller Test (ADF), H0= Model is non-stationary,
    I tried to do this based on your video but I couldn’t get stationarity in the second step what shall I do? Some pieces of literature explain that ADF test has a disadvantage for small data.
    Step 4. Determine Optimal Lag Length (k) for the model
    Step 5. Estimate VAR or VECM model
    Step 6. Perform diagnostic tests
    Is there a reference that I can use for the analysis like for example a STATA do file? The reference can be in areas of electricity or Energy Consumption and Economic Growth. I just want to understand the steps and some commands.
    Thanks, dear.

    1. Hi Ephrem,
      Sorry for the late response. Kindly reduce this. It is too long. Then post on the appropriate YouTube video and I will respond from there.
      Thanks.

  28. Good evening to everyone. Please I need a clarification on PMG. are there diagonostics tests after performing PMG estimations. if there are what tests are appropriate? Thanks

  29. Please I would like to know if there are diagonostic tests after PMG estimations. if there are what are some of them? from Ghana

    1. Helle Dr Ngozi
      I am using xtabond2 for my project. However when I do so I get a
      message from STATA that “Favoring space over speed. To switch, type or click on mata: mata set matafavor speed, perm.”. When I click on it and then run my command using xtabond2 I get the same message as above. I am perplexed why it is re appearing despite the fact that I have clicked on it in the first place. Please I need more explanation. I am using Stata14 please help me on this?

      Many thanks!

  30. Hi Crunches, please I need help. I specified ECM in Eviews 9 and always getting “VECM specification requires at least two series.” This is recurrent. What do I not get right in the specification? Again, when your series are stationary of order 1(2) or 1(1), 1(0) and 1(2) what should be done? My issue is with 1(2). Thanks

    1. Hi Oloyede,
      Sorry for the late response, but no you cannot. Next time, post on the comment section of the appropriate YouTube video and I will respond from there.
      Thanks.

  31. Dear crunch econometrix,

    I have a problem interpreting my coefficients. I use TSLS to examine the total impact of FDI on inequality in Latin America, which is a sum of 2 coefficients. These are a coefficient including all countries (also Latin American countries) and a Latin American specific coefficient. Of these two coefficients, the global one is insignificant while the Latin American coefficient is significant.
    I used the command “test _b[FDI all countries] + _b[FDI in Latin America] = 0” to assess whether the sum of the two coefficients is significant. The p-value of this test is >0.10, thus I conclude that the sum (total impact) of FDI is insignificant.

    Thus, FDI global (0.021) + FDI Latin America (-0.025***) = -0.004

    Since the sum of the coefficients (the total impact of FDI in Latin America) is insignificant, do I have to conclude that FDI has no impact on income inequality in Latin America? Or can I conclude there is some significant relationship between FDI and income inequality in Latin America as the coefficient on specific Latin American countries is significant?

    Hoping someone can help me out!

  32. I have watched your videos on gmm. Do I have to transform the data before using estabond2. Also, am finding it difficult to fix my variables to conform to your do file. My independent variable is the endogenous variable (Gender),and another independent variable is EQ, Instruments I have industry average of gender and lag of gender. DV=cost of capital and control variables are IC, SIZ, LEV, LIQ. Please help. Thanks

Leave a Reply

Your email address will not be published. Required fields are marked *