Beloved guest/subscriber, you have discovered my amazing TH-cam Channel tailored specifically for you and other beginners and intermediate users. Please do not keep me to yourself (lol). Kindly share my videos and links with your students, colleagues and academic community so that they too can SUBSCRIBE and learn with ease….and for the global community to be aware that applied econometrics can be simplified. My teaching approach is very practical. I adopt a do-as-I-do style. Many thanks to those who have supported me by telling others. Once again, CrunchEconometrix loves to teach, support my Channel with your subscription, likes, feedbacks and sharing my videos with your cohorts. Follow me on Facebook, Twitter and Reddit. Love you all, greatly!!!
Last few questions 1- can we introduce more than one lag of depedent variables in system gmm 2- can we introduce lags of independent variables in system gmm 3- should diagnostic statistics of AR2 and Hansen should remain insignificant even after introducing dummy variables or we can wink at these terms introducing the dummies
Thank you very much. Looking forward to the other videos on GMM. You do not know how these videos have saved a lot of challenges for some of us. Please keep it up.
I have no words to express how greatful I am for these videos, they have trully made my time working on the AB estimator much easier than they could have been. You are a great teacher and I hope to see more in the future!
Dear Dr. Ngozi ADELEYE. Greetings. I appreciate your tutors on GMM and I benefited more from it. Thank you very much for such informative and well prepared tutors. I have two questions for you. 1) I am doing on System GMM for my research. Does it require stationarity test? 2) If the number of instruments are greater than the observation groups, what is the impact on the analysis because in some cases the observation group may not be too large to get data. Of course my observation group is greater than the time period (N>T) but the difference is small. So, I need clarification.
Hi Taye, on your queries: 1) stationary not required. Watch my videos again. 2) read Roodman (2014, 2019) about GMM instruments for better understanding. Thanks
Dear Prof, Naturally when I watch TH-cam videos, I do so anonymously. But, when I watched your videos, honestly I felt I should sign in and commend your efforts, dedication and clarity in presenting these videos. I am glad I did so. Secondly, my concern is; I have a balanced panel data of 18 cross sections, but my time dimension is Quarterly, not annual. Thus, the time is 34 quarters or 9 years. And since the data points are quarterly and I want to estimate a dynamic panel data model, I am at loss on what to do, given that, your advice was the N>T. If it were annual data, the condition would have been met, but its quarterly data, thus N
Hi Arigisku, I'm touched by your positive feedback on my videos. Deeply appreciated! 💕 😊. On your query, N>T holds for GMM, FE and RE regardless of the nature of T, if you understand my point. So, in this case, you can't estimate GMM and neither can you estimate panel ARDL because T must be at least 30 observations. What you can do is to convert quarterly to yearly (no idea how, but search the internet) and estimate GMM. Thanks for the compliments and confidence in my teaching potentials and I hope to do better. May I know from where (location) you are reaching me?
@@CrunchEconometrix Dear Prof, My apologies, I was offline for some time. I do appreciate your reply. I will try your advice and revert back to you. As regards to your question, I am here in Nigeria, watching from Borno State. One of these days I will endeavour to visit Covenant University. In the mean time, do accept my warm regards , and please continue the good work. Stay Blessed.
@@arigiskuNo worries, at all. Kindly share my videos with your, friends, students and academic community in Nigeria and beyond and on social media for awareness. They'll learn some useful tips and skills too! :)
Dear Professor, First of all let me express my aknowlegment and gratitude for your efforts. They are keeping me afloat during these days in writing my thesis. I would like to ask you if you have any videos on how to generate lagged independent varibale from dependent varible and videos on instrumental varibles as well? If not, could you mention some sources or literature on them? Thanks in advance!
Hi Mex, thanks for the positive feedback and remarks on my videos. Deeply appreciated! I have no ideas as regards to your query. Please post it on Statalist.org for more constructive feedback. May I know from where (location) you are reaching me?
U're welcome, Abdul...kindly return the gesture by sharing my videos with your colleagues and students on social media. They need to know that econometrics can be simplified (lol)...gracias!!!
Thank you very much. You have clarified many things for me. Including the discussion about the equation(level) and if it’s necessary to specify or not. I have one question though: when should one include equation(diff). I often see this for the gmm instruments. What would be the difference between xtabond2 y L.y k w, robust small gmm(y , lag (2 3) equation (diff)) iv(k w, equation (level) and the above specification without the (diff) option? Thank you very much.
Hi Dana, xtabond2 is a user-written which has different specification. I have no idea about that spec and I will advise that you use the spec that you understand.
@@CrunchEconometrix Thank you very much. Since you were using the xtabond2 command without the gmmstyle equation(diff) option, I assumed there was a reasoning for that . But if I understand correctly you don’t know? Since without the option both equation(diff) and equation(level) are applied if I understand correctly
U're welcome, Patrick...kindly return the gesture by sharing my videos with your colleagues and students on social media. They need to know that econometrics can be simplified (lol)...gracias!!!
Hello Dr. Please i have a panel data of about forty(40) countries in sub-Saharan Africa. I have some missing values. A search gives an option to use averages to fill the gaps. kindly advise me. my email is akebapatrick@gmail.com
Hello, This is Lambert from Kigali, Rwanda. I love your teaching videos on Econometrics. I noticed that in simulation (4), the equation is not fully written, since at the end, y* [year dummy variable] is followed by comma. I was expecting to see the full model without equation () sub option, but having the collapse option. I am not sure if I am correct. I see that there is some error in writing the full model [Command].
Hi Kwizera, thanks for your positive feedback. Deeply appreciated 🙏. There are several ways to write the GMM syntax, and the one I used is complete and correct. Otherwise, I won't have been able to execute the command.
I have watched your videos on youtube, first I am very thankful to you and appreciate your work. Currently I am working on Chinese companies data. I have a unbalanced panel data which have issue of heterogenity so I used GLS method and got the results by following commands of 5 equations models: independent variable= rdr, dependent variable= tmtcd , control variables= sofln profp levp ageln grwln soenot. Moderating variables=stock and tmtgdp m1 = tmtcd * stock m2= tmtcd * tmtgdp Now I want to test my results by system GMM for robust analysis. Kindly help me to write stata code for system GMM for above variables. Thanks for cooperation.
Very great video!! very helpful to my dissertation. I want to ask to put LOG in the dependent, independent, and control variable is that necessary? because my variables are not in a log model
Dear Professor, thank you for your great video. I have learned a lot from it. I hope you can spend some time answering my questions. I am working on a firm-level panel dataset from multiple industries. Do I need to add industry dummy variables in the instrument? Besides, my result from one-step system GMM says the number of instruments is greater than the number of groups, how should I understand this and is there any way to fix this? Thank you.
Hi Vu, you may add dummy variables directly to the model and not necessarily as instruments. Kindly watch the rest of the 9 GMM videos to know what to do with too many instruments. Thanks.
Thank you so much for your great educational videos. But I still have a few questions in my mind. 1. Do I need to do something if coefficients are too small? 2. How to add more lags (>1)? Thanks a lot.
Hi Sra, thanks for the positive feedback. Deeply appreciated! I don't understand what you mean by "too small". There are are several ways to increase lags depending on what. Is it on a variable, gmm() or iv()?
Dear Professor, your videos are very helpful for me and I am very much grateful to you for sharing these videos. Ma'am my data set is having 11 year data having 90 companies. I want to know short run and long run relationship between stock return and independent variables related to economy industry and company financial ratio. But my data set some variables are stationary at level and some at level 1 and also facing the problem of autocorrelation and heteroscedasticity. Please suggest can I use GMM.
Hi Shanika, thanks for the positive feedback. Watch the foundational GMM video for proper understanding of the rudiments. Note: GMM is applicable when N > T.
Hi Festus, thanks for the positive feedback on my videos, deeply appreciated! 💕 😊. Once I'm familiar with the procedure, I'll do my best to make videos on panel cointegration techniques. May I know from where (location) you are reaching me?
@@festusndidibuogwuopara8634 Gr8! Kindly share my TH-cam Channel link and videos with your students and colleagues for awareness. A lot of them will find these tutorials very helpful. Thanks 😊
Thank you very much for your videos. I have one query, I observed that your data have some missing observation, still 'xtset' is showing strongly balanced panel. How to handle it
Thank you so much! It's a really helpful video. I wanna use GMM to do robustness test of my basic regression which is a short panel FE regression. But I have a problem using "xtabond2" to do onestep dif-GMM. the y(t-1)is not significant, but my x is significant as my basic regression, and all the AR(1) AR(2）Hansen test were past. Can I still use this result to be my robustness test? Looking forward your reply.
Hello professor, thank you very much for your videos. They're very helpfull. I only have one question. How is it that we consider the hansen test 'not problematic' even if the pvalue is never smaller than 0.05? I have understand that a pvalue of 0.4 for Hansen is ok, but I don't get why. Thank you !
Very helpful videos. Appreciate it. Just one concern I want to mention. According to Roodman’s ‘how to do xtabond2’ article, we have to use lag 2 and more for endogenous, lag 1 and more for predetermined variables. However in simulation 5 you included lag 1 for endogenous variable lngdp and lag zero for predetermined variable lnmob. Is there any background for that?
Hi Professor, Thank you for these videos, they have been very helpful! I had a few questions, if you don’t mind: 1) If a regressor is endogenous rather than predetermined (as you showed in this video), is it still okay to just put it in the gmm() instrument set? 2) What are the issues with running xtabond2 when N
Thanks Noah, I am also glad that you find the content very helpful. To your queries: (1) YES; (2) xtabond2 is not applicable. Watch my panel ARDL videos; (3) no idea. You may need to check more online resources for that. May I know from where (location) you are reaching me?
No problem, Jhabindra. If at all the SVAR videos will be made, they will be uploaded to my Teachable paid platform cruncheconometrix.teachable.com and not TH-cam. Thanks
Ma'am, could you please provide the rationale of selecting variables in the gmm list and iv list. What guides the selection of variables in each of these list?
@@CrunchEconometrix Thank you so much maam.. And can there be any assumption from the researcher side regarding exogeneity of certain regressors in our model?
Very helpful videos and the way you are explaining GMM is amazing. I want to get access to your website but I think it is blocked here in London. I am doing my PhD and I would appreciate if I can get access to your Stata dofile with the codes. Would you please share your email that way I can contact you. Thank you in advance.
Hi Ines, due to abuse and unethical conduct, dofiles are no longer free but available on my website upon payment. Here's the link cruncheconometrix.com.ng/shop. Thanks
Hello Prof. Ngozi, I'm a Master's student at Seoul National University and currently writing my thesis related to Public Debt and Economic Growth in SSA. I appreciate your insightful tutorials for beginners like me that are clear to understand and to follow. I've ran a GMM regression on a data set and I'm getting statistically insignificant p-values on ALL my year dummies. Is this normal or is there something I need to do? kindly let me know.
Hi Jo, thanks for the positive feedback on my TH-cam videos. Deeply appreciated! 😀 A lot of factors can be responsible: regressors, number of lags and number of instruments. Try reducing or changing your regressors, run simulations with different lag structures and observe the outcomes. However, the most important is that your coefficients are statistically significant even if the year dummies are not. May I know from where (location) you are reaching me?
@@josphatnjue1800 No worries, Jo. I'll appreciate if you can share the link to my TH-cam Channel with your friends and academic community in South Korea for awareness...thanks 😊
Not exactly, Abdul. It's up to the researcher to make assumptions on which variable is endogenous and applies the GMM estimator to correct for such. Thanks for watching my videos kindly share with your friends and academic community...gracias! 💕 😊 🎂
Hi! Thank you for sharing your knowledge. I would like to ask you two questions if you don't mind. Can this procedure control the non-observable heterogeneity? I ask you this because it is not clear how this matter is considered in the estimates. Second, how can I introduce an endogenous regressor in the model? For example, a demand function has 2 endogenous variables: quantity and price.
Thanks for the positive feedback...appreciated! I covered the nitty-gritty of the GMM which addresses some of your queries in my introductory video. You may need to watch that clip and also go through the suggested references listed at the end of each GMM video for in-depth knowledge about the technique. Regards.
Hi, can I know whether the Year Dummies (Y1, Y2, Y3, etc.) are required in the GMM estimation? Can I just use the Year Trend (Year) in the GMM estimation? Thank you.
Dear Professor, I've been continously watching your videos. They're super helpful. Thank you very much. However, I have a question for you. For my thesis I want to see the static model then I regressed with two way fixed effects. The coefficient of my variable of interest is positive according to the static model. After that, for the dynamic model, I included lagged-1 of dependent variable and regressed using system GMM following your steps. Yet the coefficient of my variable of interest becomes negative but still significant. I have tried to replace the variable with a proxy but the results remained. How can I interpret this? Does it mean something bad? My sample includes 88 countries covering 38 years period. I really hope you could help me. Thank you so much.
Hi Zahrina, please understand that signs of coefficients can change based on: estimation techniques, other regressors used etc. So, interpret what you have and let the readership know what the signs differ. Thanks 😊.
Hi dear, Thank you for sharing, I have a question please, Is it important that the lag GDP be significant? because I have got it not significant. Thanks a lot
Hi Sara, most of the time the lagged depvar is usually statistically significant. But since studies and scope differ that expected outcome should not be generalized. So, interpret what you have.
@@CrunchEconometrix dear mam, could you please suggest me which method I could use if my T>N? as well, i have seen some studt which used first difference GMM when T>N. Can I use it?
Hi Abdifatah, thanks for the encouraging feedback. Deeply appreciated 🙏. Please know that due to abuse and unethical conduct Stata dofiles used in my videos are no longer free but available on my website upon payment. Here's the link cruncheconometrix.com/view/datashop.php
And in iv( ) set of instruments I used 2 or 3 instruments that are not in model but I have in my dataset....according to roodman(2009) we could used variables as instruments which are not in our model...so only one should be used or we can add more than one to improve our results
Thank you very much for the videos, Prof. They are really helpful, you are doing a great job. I am currently running a regression using SGMM. I have N=15 and T=4 (years average). I am assessing the effect of agricultural aid on poverty headcount ratio. Stata gives the following: Arellano-Bond test for AR(2) in first differences: z = . Pr > z = . It's just a dot. What does it mean? How can I fix that?
Hi Kan, several reasons for that: insufficient observations, too many instruments, or too many missing values in the dataset. May I know from where (location) you are reaching me?
@@CrunchEconometrix Thank you for your reply. I am reaching you from Ivory Coast. What can I do in this case? Should I increase my sample? The issue is the data on poverty ratio are not available on a long period, and if I increase the sample, I may fail to meet the requirement that N>T. So far, N=15 (15 ECOWAS countries) and T=4.
@@kandavidndri5938 Yeah, inequality and poverty data often have those missing values characteristics, particularly for Africa. You can increase the time span to 12 years if there are observations and still maintain N>T. I will appreciate if you can share my YT Channel link with your student, friends and academic community in Cote d'Ivoire...thanks!
I've increased the number of years, and now the AR(2) test appears, but the Hansen test does not...Can I take it like that or is there something wrong again?
Thanks for this video professor. When I use the collapse option, Stata does not run my results and issues this message "Equation not identified. Regressors outnumber instruments",. I have a time period of 23 years with dummies created and added 4 predetermined instrumental variables. What do I do I do not also understand how Stata calculates and report that "Number of instruments = 257"
Dear Prof, I am wondering how can I decide which variables belong to gmmstyle() and which variables belong to ivstyle(). I know that varaibles in iv() should be exogenous, but how can I know it? Best wishes.
want to ask why sargan test (using 2 step gmm estimator) is always resulting in p value equals to 1 while p value of one step gmm is varying. cant we use sargon test for two step gmm. Thanks in Advance
@@CrunchEconometrix thank you so much prof, i was unable to define what is gmm. after watching your video now i m in the growth stage and easily understand the gmm. Q1. Any video regarding the dummy for panel data coz, my obs, is 64, Q1. any video regarding the panel data analysis ???? thank you so much for your effort cant express in words,,, God blesse you
Thanks Asghar, for the encouraging feedback. Deeply appreciated! Kindly watch my videos on "Error Component Models" in addition to the videos on "Dummy Variables".
Thank you for your for your very insightful videos, I am a new subscriber, my SGMM model has 13 cross sectional data and 9 observations(5-years moving average data), however, I still have series of problems with stata, simply put, stata gives me error r100(no observation) all the time. My concern is, how can I do this analysis using SGMM and stata?
That means you should have about 112 observation. However, a "no obs" error can also occur if one of your variables does not have data points. Check your data again...
@@CrunchEconometrix Thank you very much, I think the problem is that STATA identifies my data as unbalanced with GAPs. hence, I think the no observation error am getting is probably because there are gaps with the times i.e( 1974, 1978, 1982..... 2014). what do you suggest I do ma'am?
@@CrunchEconometrix Let r be the total number of moment functions, q be the number of variables, so J-test follows Chi-square at the degree of freedom of r-q. Thus I am wondering how to get the degree of freedom. In your example, I noticed the degree of freedom was changed when you used different numbers of groups. So how can I calculate the degree of freedom in detail?
As mentioned, consider using a PMG, MG or DFE model. Beware of the number of variables to include in the model because of the problem of the curse of dimensionality
U're welcome, Nur...kindly return the gesture by sharing my videos with your colleagues and students on social media. They need to know that econometrics can be simplified (lol)...gracias!!!
You have to perform the different simulations just like I did and watch out for the value of AR(2) as you adjust the instruments set and/or regressors.
Hi Suleiman, I deactivated the link due to abuse and unethical conduct. There were several attempts to hack my Google drive. Since August 2019, all Stata dofiles used in my videos and some datasets are no longer free but available on my website upon payment. Here's the link cruncheconometrix.com/view/datashop.php
How much lags of endogenous variable could be used in gmm style instruments gmm ( ) you just take first lag....only 1st lag is used or we could used 2 or 3 lags of endogenous variables to improve our results?
You can use as many lags your model can take. But check the signs and statistical significance of your variables, AR(2), and Hausman statistics as you go along. Personally, I have used up to 3 lags in gmm( ).
@@CrunchEconometrix Thankxxx You to solve my queries .....Your videos help me alot on the other side your kind suggestions solve my problems..stay blessed
Hi professor Only ON step System gMM is working for my data set.But IF i use robust option then the result isn't good.Only if i use without robust then ,my result is good.But may i report it one step gmm result without robust option (without hansen result also as without robust ,there is no hansen result) in my thesis?Can you please say something?I am in so much problem.
THANK YOU PROFESSOR. PROF, i WOULD LIKE TO ASK JUST ON SPECIFIC THINGS. May I REPORT MY ONE STEP GMM RESULT WITHOUT HANSEN TEST IN MY MASTERS THESIS? IS IT ACCEPTABLE? @@CrunchEconometrix
Thank you Professor I always Follow You Prof.In my thesis,I am estimating the impact of money laundering on economic growth of developing countries. Now i need to extend a growth theory But i can not connect with any growth theory.I want to extend solow growth theory but what can be the logic behind considering this theory.How can i connect money laundering with solow growth theory?Can you give me a way if possible /? I am in trouble and about to be dropped out from program. Thank you Prof. I am also showing the moderation impact of money laundering between foreign direct investment and economic growth and between financial sector development and economic growth. @@CrunchEconometrix
@@CrunchEconometrix thank you. If my AR 2 is rejected and Hansen value is .224 whereas number of instruments are 13 and groups are 150. With lag (. 2) can I say the model is good.
@@CrunchEconometrix Thank you very much for response. AR(2) is statistically insignificant. As you mentioned and in some papers it is in line with the specifications of a good model.
Gmm is so confusing method I have different results between e views and stata. Honestly I don't have much trust towards it. Just one question I had better results when removing year dummies. What do you think?
Hi Abdelhadi, yes you are likely to get different results due to different programming languages. If your results are better without year dummies go ahead and estimate but make a note in your results table stating why year dummies are excluded.
Hi Sal, due to abuse and unethical conduct Stata dofiles used in my videos are no longer free but available on my website upon payment. Here's the link cruncheconometrix.com/view/datashop.php You can install the "xtabond2" syntax on Stata by using either "ssc install xtabond2" OR "help xtabond2" OR "findit xtabond2".
Thanks for the reply. When I try to install the xtabond2 command using the ssc install xtabond2 command, the following message appears: ssc install xtabond2 checking xtabond2 consistency and verifying not already installed... the following files already exist and are different: c:\ado\plus\l\lxtabond2.mlib@@CrunchEconometrix
@@CrunchEconometrix have tried multiple options but I am continuously getting this result of hansen test ....Hansen test of overid chi2(15) = 0.00 Prob > chi2 = 1.000. can you suggest where would the issue be in my analysis or results?
![“อ.เบียร์ คนตื่นธรรม” มาตามคำเรียกร้อง แหกศาสตร์ฮวงจุ้ย เตือนสติอย่างมงาย | แฉ 7 พ.ย. 67 [1/3]](http://i.ytimg.com/vi/5fc4S3xBNfc/mqdefault.jpg)
Beloved guest/subscriber, you have discovered my amazing TH-cam Channel tailored specifically for you and other beginners and intermediate users. Please do not keep me to yourself (lol). Kindly share my videos and links with your students, colleagues and academic community so that they too can SUBSCRIBE and learn with ease….and for the global community to be aware that applied econometrics can be simplified. My teaching approach is very practical. I adopt a do-as-I-do style. Many thanks to those who have supported me by telling others. Once again, CrunchEconometrix loves to teach, support my Channel with your subscription, likes, feedbacks and sharing my videos with your cohorts. Follow me on Facebook, Twitter and Reddit. Love you all, greatly!!!
Kindly sent me text file about all your videos
@@sarfrazhussain9851 If you mean the dofiles, they are available on my website. Link is at the end of the video.
Last few questions
1- can we introduce more than one lag of depedent variables in system gmm
2- can we introduce lags of independent variables in system gmm
3- should diagnostic statistics of AR2 and Hansen should remain insignificant even after introducing dummy variables or we can wink at these terms introducing the dummies
Thank you very much. Looking forward to the other videos on GMM. You do not know how these videos have saved a lot of challenges for some of us. Please keep it up.
Thanks so much for this feedback Abel, gr8 to know that these clips have been helpful....kindly tell others about my Channel! 💕 😊
I have no words to express how greatful I am for these videos, they have trully made my time working on the AB estimator much easier than they could have been. You are a great teacher and I hope to see more in the future!
Great to hear, Nico...thanks for the positive feedback!
Dear Dr. Ngozi ADELEYE. Greetings. I appreciate your tutors on GMM and I benefited more from it. Thank you very much for such informative and well prepared tutors. I have two questions for you. 1) I am doing on System GMM for my research. Does it require stationarity test? 2) If the number of instruments are greater than the observation groups, what is the impact on the analysis because in some cases the observation group may not be too large to get data. Of course my observation group is greater than the time period (N>T) but the difference is small. So, I need clarification.
Hi Taye, on your queries:
1) stationary not required. Watch my videos again.
2) read Roodman (2014, 2019) about GMM instruments for better understanding. Thanks
Dear Prof,
Naturally when I watch TH-cam videos, I do so anonymously. But, when I watched your videos, honestly I felt I should sign in and commend your efforts, dedication and clarity in presenting these videos. I am glad I did so.
Secondly, my concern is; I have a balanced panel data of 18 cross sections, but my time dimension is Quarterly, not annual. Thus, the time is 34 quarters or 9 years. And since the data points are quarterly and I want to estimate a dynamic panel data model, I am at loss on what to do, given that, your advice was the N>T. If it were annual data, the condition would have been met, but its quarterly data, thus N
Hi Arigisku, I'm touched by your positive feedback on my videos. Deeply appreciated! 💕 😊. On your query, N>T holds for GMM, FE and RE regardless of the nature of T, if you understand my point. So, in this case, you can't estimate GMM and neither can you estimate panel ARDL because T must be at least 30 observations. What you can do is to convert quarterly to yearly (no idea how, but search the internet) and estimate GMM. Thanks for the compliments and confidence in my teaching potentials and I hope to do better. May I know from where (location) you are reaching me?
@@CrunchEconometrix Dear Prof,
My apologies, I was offline for some time. I do appreciate your reply. I will try your advice and revert back to you. As regards to your question, I am here in Nigeria, watching from Borno State. One of these days I will endeavour to visit Covenant University. In the mean time, do accept my warm regards , and please continue the good work. Stay Blessed.
@@arigiskuNo worries, at all. Kindly share my videos with your, friends, students and academic community in Nigeria and beyond and on social media for awareness. They'll learn some useful tips and skills too! :)
Really good explanation! Thank you very much
U're welcome, Alex...kindly tell others! 💕 😊
Dear Professor,
First of all let me express my aknowlegment and gratitude for your efforts. They are keeping me afloat during these days in writing my thesis. I would like to ask you if you have any videos on how to generate lagged independent varibale from dependent varible and videos on instrumental varibles as well? If not, could you mention some sources or literature on them?
Thanks in advance!
Hi Mex, thanks for the positive feedback and remarks on my videos. Deeply appreciated! I have no ideas as regards to your query. Please post it on Statalist.org for more constructive feedback. May I know from where (location) you are reaching me?
Thank you so much sour explanations are absolutely easy to understand
Thanks a lot for the sustained effort. Your contribution has been helpful.
U're welcome, Abdul...kindly return the gesture by sharing my videos with your colleagues and students on social media. They need to know that econometrics can be simplified (lol)...gracias!!!
Thank you very much. You have clarified many things for me. Including the discussion about the equation(level) and if it’s necessary to specify or not. I have one question though: when should one include equation(diff). I often see this for the gmm instruments.
What would be the difference between xtabond2 y L.y k w, robust small gmm(y , lag (2 3) equation (diff)) iv(k w, equation (level) and the above specification without the (diff) option?
Thank you very much.
Hi Dana, xtabond2 is a user-written which has different specification. I have no idea about that spec and I will advise that you use the spec that you understand.
@@CrunchEconometrix
Thank you very much. Since you were using the xtabond2 command without the gmmstyle equation(diff) option, I assumed there was a reasoning for that . But if I understand correctly you don’t know? Since without the option both equation(diff) and equation(level) are applied if I understand correctly
What are you saying?
Thanks very much Dr. most grateful for all your videos. Kindly do this with e views. Thanks.
U're welcome, Patrick...kindly return the gesture by sharing my videos with your colleagues and students on social media. They need to know that econometrics can be simplified (lol)...gracias!!!
Hello Dr. Please i have a panel data of about forty(40) countries in sub-Saharan Africa. I have some missing values. A search gives an option to use averages to fill the gaps. kindly advise me. my email is akebapatrick@gmail.com
Hello, This is Lambert from Kigali, Rwanda. I love your teaching videos on Econometrics. I noticed that in simulation (4), the equation is not fully written, since at the end, y* [year dummy variable] is followed by comma. I was expecting to see the full model without equation () sub option, but having the collapse option. I am not sure if I am correct. I see that there is some error in writing the full model [Command].
Hi Kwizera, thanks for your positive feedback. Deeply appreciated 🙏. There are several ways to write the GMM syntax, and the one I used is complete and correct. Otherwise, I won't have been able to execute the command.
I have watched your videos on youtube, first I am very thankful to you and appreciate your work. Currently I am working on Chinese companies data. I have a unbalanced panel data which have issue of heterogenity so I used GLS method and got the results by following commands of 5 equations models:
independent variable= rdr, dependent variable= tmtcd , control variables= sofln profp levp ageln grwln soenot. Moderating variables=stock and tmtgdp
m1 = tmtcd * stock
m2= tmtcd * tmtgdp
Now I want to test my results by system GMM for robust analysis. Kindly help me to write stata code for system GMM for above variables. Thanks for cooperation.
Hi Muhd, thanks for the feedback on my videos. Deeply appreciated. But I'd responded to your email on this. Stata dofiles are available on my website.
good explanation! Thank you very much
U're welcome Lamar...kindly share my videos with your academic and social media community! :)
Very great video!! very helpful to my dissertation. I want to ask to put LOG in the dependent, independent, and control variable is that necessary? because my variables are not in a log model
Hi there, using the log of variables is at the discretion of the researcher.
Dear Professor, thank you for your great video. I have learned a lot from it. I hope you can spend some time answering my questions. I am working on a firm-level panel dataset from multiple industries. Do I need to add industry dummy variables in the instrument? Besides, my result from one-step system GMM says the number of instruments is greater than the number of groups, how should I understand this and is there any way to fix this? Thank you.
Hi Vu, you may add dummy variables directly to the model and not necessarily as instruments. Kindly watch the rest of the 9 GMM videos to know what to do with too many instruments. Thanks.
Thank you so much for your great educational videos. But I still have a few questions in my mind.
1. Do I need to do something if coefficients are too small?
2. How to add more lags (>1)?
Thanks a lot.
Hi Sra, thanks for the positive feedback. Deeply appreciated! I don't understand what you mean by "too small". There are are several ways to increase lags depending on what. Is it on a variable, gmm() or iv()?
Dear Professor, your videos are very helpful for me and I am very much grateful to you for sharing these videos. Ma'am my data set is having 11 year data having 90 companies. I want to know short run and long run relationship between stock return and independent variables related to economy industry and company financial ratio. But my data set some variables are stationary at level and some at level 1 and also facing the problem of autocorrelation and heteroscedasticity. Please suggest can I use GMM.
Hi Nishu, having watched this video and gone through the references listed at the end of the clip, you should know that GMM is applicable.
Thank you so much ma'am for increasing my confidence
Thank you very much for clear explanation. My data set consist with only 6 countries information. Is it enough to use one-step System GMM?
Hi Shanika, thanks for the positive feedback. Watch the foundational GMM video for proper understanding of the rudiments. Note: GMM is applicable when N > T.
@@CrunchEconometrix Yes Thank you. I got an idea after looking on it
Thanks for your work. Really educative. Could you please prepare a lecture on panel cointegration techniques
Hi Festus, thanks for the positive feedback on my videos, deeply appreciated! 💕 😊. Once I'm familiar with the procedure, I'll do my best to make videos on panel cointegration techniques. May I know from where (location) you are reaching me?
@@CrunchEconometrix Thanks for your prompt response. I am an Economics PhD student of Delta State University, Abraka, Nigeria
@@festusndidibuogwuopara8634 Gr8! Kindly share my TH-cam Channel link and videos with your students and colleagues for awareness. A lot of them will find these tutorials very helpful. Thanks 😊
Thank you very much for your videos. I have one query, I observed that your data have some missing observation, still 'xtset' is showing strongly balanced panel. How to handle it
Swarna, that's the response from Stata...I simply go with it and proceed with my analysis.
Thank you ma'am for quick reply, then how we will instruct stata that we have missing observations.
Sharma, I only do "xtset". Maybe there's another command that goes with it I wouldn't know, honestly.
@@CrunchEconometrix Okay ma'am is there any issues if we run the model without declaring the missing observation, the way you did
Swarna, estimate your model.
Thank you so much! It's a really helpful video. I wanna use GMM to do robustness test of my basic regression which is a short panel FE regression. But I have a problem using "xtabond2" to do onestep dif-GMM. the y(t-1)is not significant, but my x is significant as my basic regression, and all the AR(1) AR(2）Hansen test were past. Can I still use this result to be my robustness test? Looking forward your reply.
Hi Yanan, yes you can. That the coeff of the lag is not statistically significant does not invalidate your result.
Hello professor, thank you very much for your videos. They're very helpfull. I only have one question. How is it that we consider the hansen test 'not problematic' even if the pvalue is never smaller than 0.05? I have understand that a pvalue of 0.4 for Hansen is ok, but I don't get why.
Thank you !
Giulia, I'd refer you to read Roodman (2009, 2014) who made the assertion.
Thank you for the videos!! What would be the result if T > N? Even if just by a very small number?
To the best of my knowledge, this variant of GMM is only applicable to N>T panel structure.
Thanks alot. do u have a GMM test using Eviews, if so please I need the link. Thanks
Hi Khaled, thanks for the positive feedback. My GMM videos are based on Stata.
Thanks for the helpful video. May I ask why we need the year dummies? Can we skip this?
Dummy variables (DVs) are used for several reasons. I have videos on that...and no, it is not compulsory to add DVs into the GMM model.
Very helpful videos. Appreciate it. Just one concern I want to mention. According to Roodman’s ‘how to do xtabond2’ article, we have to use lag 2 and more for endogenous, lag 1 and more for predetermined variables. However in simulation 5 you included lag 1 for endogenous variable lngdp and lag zero for predetermined variable lnmob. Is there any background for that?
I'm not aware of any strict lag numbers for xtabond2.
Hi Professor,
Thank you for these videos, they have been very helpful! I had a few questions, if you don’t mind:
1) If a regressor is endogenous rather than predetermined (as you showed in this video), is it still okay to just put it in the gmm() instrument set?
2) What are the issues with running xtabond2 when N
Thanks Noah, I am also glad that you find the content very helpful. To your queries: (1) YES; (2) xtabond2 is not applicable. Watch my panel ARDL videos; (3) no idea. You may need to check more online resources for that. May I know from where (location) you are reaching me?
@@CrunchEconometrix Thank you! Yes I've been trying to run xtabond2 on a N
@@noahspencer9188 So true, Noah! Please let tell Canada 🍁 to subscribe to my TH-cam Channel! 😀
Waiting for video of SVAR model
No problem, Jhabindra. If at all the SVAR videos will be made, they will be uploaded to my Teachable paid platform cruncheconometrix.teachable.com and not TH-cam. Thanks
Dear Professor, can you have a coefficient next to lagged dependant variable that is above 1 in system GMM?
Do you mean if the coefficient of the lagged depvar is > 1?
Ma'am, could you please provide the rationale of selecting variables in the gmm list and iv list. What guides the selection of variables in each of these list?
Hi Rochna, theory, related studies and intuition are some of the factors guiding variables selection.
@@CrunchEconometrix Thank you so much maam..
And can there be any assumption from the researcher side regarding exogeneity of certain regressors in our model?
Yes, you can make your own asset.
Very helpful videos and the way you are explaining GMM is amazing. I want to get access to your website but I think it is blocked here in London. I am doing my PhD and I would appreciate if I can get access to your Stata dofile with the codes. Would you please share your email that way I can contact you. Thank you in advance.
Hi Ines, due to abuse and unethical conduct, dofiles are no longer free but available on my website upon payment. Here's the link cruncheconometrix.com.ng/shop. Thanks
@@CrunchEconometrix Thank you. the problem is that the website is blocked and I can not get access.
Ok. Ines, let me verify this and get back to you. Thanks
@@inessaddi784 Please try again. Problem resolved. Thanks.
Hello Prof. Ngozi,
I'm a Master's student at Seoul National University and currently writing my thesis related to Public Debt and Economic Growth in SSA. I appreciate your insightful tutorials for beginners like me that are clear to understand and to follow. I've ran a GMM regression on a data set and I'm getting statistically insignificant p-values on ALL my year dummies. Is this normal or is there something I need to do? kindly let me know.
Hi Jo, thanks for the positive feedback on my TH-cam videos. Deeply appreciated! 😀 A lot of factors can be responsible: regressors, number of lags and number of instruments. Try reducing or changing your regressors, run simulations with different lag structures and observe the outcomes. However, the most important is that your coefficients are statistically significant even if the year dummies are not. May I know from where (location) you are reaching me?
@@CrunchEconometrix Thank you for your quick response, much appreciated. I'm reaching you from Seoul, Korea.
@@josphatnjue1800 No worries, Jo. I'll appreciate if you can share the link to my TH-cam Channel with your friends and academic community in South Korea for awareness...thanks 😊
@@CrunchEconometrix Most certainly Prof. I will share. Regards
Thank you, Professor.
U're welcome, Moin...please share my videos with your friends and academic community. They need to know...gracias! 💕 😊
Do we need to test for endogeneity to identify endogenous variables? Thanks for your nice video.
Not exactly, Abdul. It's up to the researcher to make assumptions on which variable is endogenous and applies the GMM estimator to correct for such. Thanks for watching my videos kindly share with your friends and academic community...gracias! 💕 😊 🎂
Hi! Thank you for sharing your knowledge. I would like to ask you two questions if you don't mind.
Can this procedure control the non-observable heterogeneity? I ask you this because it is not clear how this matter is considered in the estimates.
Second, how can I introduce an endogenous regressor in the model? For example, a demand function has 2 endogenous variables: quantity and price.
Thanks for the positive feedback...appreciated! I covered the nitty-gritty of the GMM which addresses some of your queries in my introductory video. You may need to watch that clip and also go through the suggested references listed at the end of each GMM video for in-depth knowledge about the technique. Regards.
Hi, can I know whether the Year Dummies (Y1, Y2, Y3, etc.) are required in the GMM estimation? Can I just use the Year Trend (Year) in the GMM estimation? Thank you.
Hi, either approach is fine. You can exclude year dummies if using them worsens your results.
@@CrunchEconometrix Thank you for the reply. And, thank you for sharing the knowledge!
Dear Professor, I've been continously watching your videos. They're super helpful. Thank you very much. However, I have a question for you. For my thesis I want to see the static model then I regressed with two way fixed effects. The coefficient of my variable of interest is positive according to the static model. After that, for the dynamic model, I included lagged-1 of dependent variable and regressed using system GMM following your steps. Yet the coefficient of my variable of interest becomes negative but still significant. I have tried to replace the variable with a proxy but the results remained. How can I interpret this? Does it mean something bad? My sample includes 88 countries covering 38 years period. I really hope you could help me. Thank you so much.
Hi Zahrina, please understand that signs of coefficients can change based on: estimation techniques, other regressors used etc. So, interpret what you have and let the readership know what the signs differ. Thanks 😊.
Dear Madam, should lags dependant variable be insignificant.
Hi Ubaid, in most cases they should.
Hi dear, Thank you for sharing, I have a question please, Is it important that the lag GDP be significant? because I have got it not significant. Thanks a lot
Hi Sara, most of the time the lagged depvar is usually statistically significant. But since studies and scope differ that expected outcome should not be generalized. So, interpret what you have.
What should be the best fit value of Ar(1) Ar(2) and sargan test?? I'm not getting this point Ma'am
Please watch the introductory video before watching the rest of the GMM series. I gave detailed explanation.
Thank you so much mam. Mam, I have one doubt .which method I could use if number of countries less than years ?please
Hi Munshir, the xtabond GMM is for N > T panels.
@@CrunchEconometrix thank u so much mam
@@CrunchEconometrix dear mam, can I, please, get your email id ,i had some doubts to ask regarding my data analysis
Hi Munshir, kindly post your query here, and I will do my best to guide you.
@@CrunchEconometrix dear mam, could you please suggest me which method I could use if my T>N? as well, i have seen some studt which used first difference GMM when T>N. Can I use it?
Dear Lecturer, first thanks for sharing this important lesson for us, could you share the do file with us
Hi Abdifatah, thanks for the encouraging feedback. Deeply appreciated 🙏. Please know that due to abuse and unethical conduct Stata dofiles used in my videos are no longer free but available on my website upon payment. Here's the link cruncheconometrix.com/view/datashop.php
And in iv( ) set of instruments I used 2 or 3 instruments that are not in model but I have in my dataset....according to roodman(2009) we could used variables as instruments which are not in our model...so only one should be used or we can add more than one to improve our results
Yes, that is correct.
@@CrunchEconometrix more than one instruments could be used?which are not in our model?
Yes.
@@CrunchEconometrix thnkxx
Thank you very much for the videos, Prof. They are really helpful, you are doing a great job. I am currently running a regression using SGMM. I have N=15 and T=4 (years average). I am assessing the effect of agricultural aid on poverty headcount ratio. Stata gives the following:
Arellano-Bond test for AR(2) in first differences: z = . Pr > z = .
It's just a dot. What does it mean? How can I fix that?
Hi Kan, several reasons for that: insufficient observations, too many instruments, or too many missing values in the dataset. May I know from where (location) you are reaching me?
@@CrunchEconometrix Thank you for your reply. I am reaching you from Ivory Coast. What can I do in this case? Should I increase my sample? The issue is the data on poverty ratio are not available on a long period, and if I increase the sample, I may fail to meet the requirement that N>T. So far, N=15 (15 ECOWAS countries) and T=4.
@@kandavidndri5938 Yeah, inequality and poverty data often have those missing values characteristics, particularly for Africa. You can increase the time span to 12 years if there are observations and still maintain N>T. I will appreciate if you can share my YT Channel link with your student, friends and academic community in Cote d'Ivoire...thanks!
@@CrunchEconometrix Sure, I'll share your YT channel. Thank you so very much.
I've increased the number of years, and now the AR(2) test appears, but the Hansen test does not...Can I take it like that or is there something wrong again?
Is it possible to add the lag of the variable after the gmm (as endogenous) and the same variable (but in the original version) after the iv command?
Hi Barhom, I responded to you on this on another thread. The answer is NO.
@@CrunchEconometrix Thanks 🙏🙏
Thanks for this video professor. When I use the collapse option, Stata does not run my results and issues this message
"Equation not identified. Regressors outnumber instruments",. I have a time period of 23 years with dummies created and added 4 predetermined instrumental variables. What do I do
I do not also understand how Stata calculates and report that "Number of instruments = 257"
Hi Beri, Stata error message is clear. You have too many regressors. Reduce them.
hellow sir, can i use GMM estimation ? my data set is including 18 years and 5 countries
Hi Saftain, you can't. Please watch my videos on panel ARDL. Thanks.
Dear Prof,
I am wondering how can I decide which variables belong to gmmstyle() and which variables belong to ivstyle(). I know that varaibles in iv() should be exogenous, but how can I know it?
Best wishes.
Read papers and align with theory and similar studies to inform your classification. Please may I know from where (location) you are reaching me?
why was infl not in the instrument? What does it mean when you do that, for those variables not in the instrument?
Hi Enk, I gave clear explanations on every simulation and their outcomes. Kindly watch the entire GMM series for clearer understanding.
Do I need to remove multicollinearity problem before running GMM?
Of course...from the correlation analysis you can detect this so you don't include highly collinear variables together in the model.
want to ask why sargan test (using 2 step gmm estimator) is always resulting in p value equals to 1 while p value of one step gmm is varying. cant we use sargon test for two step gmm. Thanks in Advance
Hi Zaira, I explained why in the introductory GMM video. Kindly watch it, thanks.
@@CrunchEconometrix thank you so much prof, i was unable to define what is gmm. after watching your video now i m in the growth stage and easily understand the gmm.
Q1. Any video regarding the dummy for panel data coz, my obs, is 64,
Q1. any video regarding the panel data analysis ????
thank you so much for your effort cant express in words,,, God blesse you
Thanks Asghar, for the encouraging feedback. Deeply appreciated! Kindly watch my videos on "Error Component Models" in addition to the videos on "Dummy Variables".
@@CrunchEconometrix thanks you prof, highly encourage we society need such type and helpful teacher. If you was my PhD supervisor we learn from u
Is there any video about Times Series GMM !
No
Thank you for your for your very insightful videos, I am a new subscriber, my SGMM model has 13 cross sectional data and 9 observations(5-years moving average data), however, I still have series of problems with stata, simply put, stata gives me error r100(no observation) all the time. My concern is, how can I do this analysis using SGMM and stata?
Hi Lokki, total observations is derived from N*T, so how do you come about 9 obs?
@@CrunchEconometrix I meant 13 groups and 9 years.. I have been a bit paranoid for two days trying to find solutions to this problem.
That means you should have about 112 observation. However, a "no obs" error can also occur if one of your variables does not have data points. Check your data again...
@@CrunchEconometrix Thank you very much, I think the problem is that STATA identifies my data as unbalanced with GAPs. hence, I think the no observation error am getting is probably because there are gaps with the times i.e( 1974, 1978, 1982..... 2014). what do you suggest I do ma'am?
Since gaps are related to your variables, so you may have to drop those affected and find better proxies with sufficient data points.
Hi, I am wondering how can I tell the total moment functions from your model?
Lei, I don't know what you mean by "total moment functions".
@@CrunchEconometrix Let r be the total number of moment functions, q be the number of variables, so J-test follows Chi-square at the degree of freedom of r-q. Thus I am wondering how to get the degree of freedom. In your example, I noticed the degree of freedom was changed when you used different numbers of groups. So how can I calculate the degree of freedom in detail?
I have no idea, Lei. You may want to check out other online resources.
Mam, if N
Piyali, kindly post this on an econometric platform for more advise. May I know from where (location) you are reaching me?
As mentioned, consider using a PMG, MG or DFE model. Beware of the number of variables to include in the model because of the problem of the curse of dimensionality
Sir may I please know why pvalue of AR(2) must be greater than 0.05? And AR(1) less than 0.05
Kindly read any of the GMM articles listed at the end of the video for detailed knowledge about your query. Thanks.
Thank you so much!
U're welcome, Nur...kindly return the gesture by sharing my videos with your colleagues and students on social media. They need to know that econometrics can be simplified (lol)...gracias!!!
how to cover the serial correlation problem in STATA
You have to perform the different simulations just like I did and watch out for the value of AR(2) as you adjust the instruments set and/or regressors.
What is Y * mean. Is it dummy for year?
Yes.
Hello mam, I checked your website, but couldn't find the do file
Hi Suleiman, I deactivated the link due to abuse and unethical conduct. There were several attempts to hack my Google drive. Since August 2019, all Stata dofiles used in my videos and some datasets are no longer free but available on my website upon payment. Here's the link cruncheconometrix.com/view/datashop.php
How much lags of endogenous variable could be used in gmm style instruments gmm ( ) you just take first lag....only 1st lag is used or we could used 2 or 3 lags of endogenous variables to improve our results?
You can use as many lags your model can take. But check the signs and statistical significance of your variables, AR(2), and Hausman statistics as you go along. Personally, I have used up to 3 lags in gmm( ).
@@CrunchEconometrix OK Thankxxx
@@CrunchEconometrix Thankxxx You to solve my queries .....Your videos help me alot on the other side your kind suggestions solve my problems..stay blessed
@@CrunchEconometrix Hausman or Hansen?
How I can get this text please share the link from where I can download
Hi Sarfraz, the text is not available. But you can get some notes from the video description. Thanks.
Is it mandatory to have year dummies? I am getting better results without them.
Hi Abhijit, you can estimate without year dummies.
Hi professor
Only ON step System gMM is working for my data set.But IF i use robust option then the result isn't good.Only if i use without robust then ,my result is good.But may i report it one step gmm result without robust option (without hansen result also as without robust ,there is no hansen result) in my thesis?Can you please say something?I am in so much problem.
I'd advise you use the approach that works for your dataset.
THANK YOU PROFESSOR.
PROF, i WOULD LIKE TO ASK JUST ON SPECIFIC THINGS.
May I REPORT MY ONE STEP GMM RESULT WITHOUT HANSEN TEST IN MY MASTERS THESIS?
IS IT ACCEPTABLE?
@@CrunchEconometrix
It's important you indicate the Hansen statistic.
Thank you Professor
I always Follow You
Prof.In my thesis,I am estimating the impact of money laundering on economic growth of developing countries.
Now i need to extend a growth theory But i can not connect with any growth theory.I want to extend solow growth theory but what can be the logic behind considering this theory.How can i connect money laundering with solow growth theory?Can you give me a way if possible /? I am in trouble and about to be dropped out from program.
Thank you Prof.
I am also showing the moderation impact of money laundering between foreign direct investment and economic growth and between financial sector development and economic growth.
@@CrunchEconometrix
Can you tell me the effect of changing the lag value
Salman, you will see that from your output when you compare the results.
@@CrunchEconometrix thank you. If my AR 2 is rejected and Hansen value is .224 whereas number of instruments are 13 and groups are 150. With lag (. 2) can I say the model is good.
Your model is not good. AR(2) must NOT be statistically significant.
@@CrunchEconometrix Thank you very much for response. AR(2) is statistically insignificant. As you mentioned and in some papers it is in line with the specifications of a good model.
Gmm is so confusing method I have different results between e views and stata. Honestly I don't have much trust towards it. Just one question I had better results when removing year dummies. What do you think?
Hi Abdelhadi, yes you are likely to get different results due to different programming languages. If your results are better without year dummies go ahead and estimate but make a note in your results table stating why year dummies are excluded.
HiDr. , please I want the GMM codes you made for xtabond2 so I can install it on stata 16?
Hi Sal, due to abuse and unethical conduct Stata dofiles used in my videos are no longer free but available on my website upon payment. Here's the link cruncheconometrix.com/view/datashop.php
You can install the "xtabond2" syntax on Stata by using either "ssc install xtabond2" OR "help xtabond2" OR "findit xtabond2".
Thanks for the reply. When I try to install the xtabond2 command using the ssc install xtabond2 command, the following message appears:
ssc install xtabond2
checking xtabond2 consistency and verifying not already installed...
the following files already exist and are different:
c:\ado\plus\l\lxtabond2.mlib@@CrunchEconometrix
Hansen test of overid chi2(15) = 0.00 Prob > chi2 = 1.000 where it may be my prob ??
I don't quite understand your question. Kindly rephrase. Thanks.
@@CrunchEconometrix have tried multiple options but I am continuously getting this result of hansen test ....Hansen test of overid chi2(15) = 0.00 Prob > chi2 = 1.000. can you suggest where would the issue be in my analysis or results?
You may need to redo your model. Change/reduce independent variables.
09:05 no collapse cmd
Play the clip and listen my explanations.
why Hansen-j-test p-value close to 1 isn't a good sign
If I add an instrumental variable, wouldn't the p-value invariant of the hansen j test be the best?
Kindly read Roodman (2009, 2014) for indepth explanation. Reference list at the end of the video.
I don't understand what you mean by this.