You are amazing! Your videos have helped me so much throughout both my BA and MA degree in Economics. Now I am starting my PhD and they are still helping me a lot. Thanks so much for the work you put into it! We really appreciate it
I had to skip most of undergrad and would now be struggling through the early stages of my Master's, were it not for your videos filling the gaps! Thank you.
because no one ever explains what the individual items in the equation means...such as his statement at 0:44 where he says something does not equal S...what the actual fudge is S?
@@scottyagreen The s refers to the index of Us (=error term for entity/observation s). If i =s, then the correlation equals 1 because two identitcal things are perfectly correlated. This case however is of course not relevant .
@@scottyagreen i and s are any numbers which are not the same, they are talking about the indexes. For example: when he says the model Y_i= Alpha+ BetaX_i +U_i. Here he talks about one particular sample numbered i={1,2,3,4,5}. Therefore U_i=error term for sample{1,2,3,4,5} U_s=error term for sample{1,2,3,4,5} also, but as i=/u thus when i takes the value of 1, s cannot be 1, it can be anything else{2,3,4,5}.
@@scottyagreen i and s are both subscripts indicating the ith and the sth term respectively. Example : i=1 tests the covariance of the first term If there are 3 terms - covariance of the 1st term will be tested against the 2nd and 3rd term s represents the terms it is compared against so here, s= {2,3} s shouldn't be equal to 1 as it would mean checking covariance of the 1st term (ith) with itself (sth) and that's not needed
Great vid as usual.. Would it be possible for you to do an illustration of the rank condition for simultaneous equations... I already know how to show if it is exactly identified or over identified using the 'order condition' However I really cant seem to understand the rank which i know is the necessary and sufficient condition. Regards.
hi, i have a question in a practice paper "What is the difference between autocorrelation and serial correlation? [2 marks] " is the answer that there is no difference?
Hi, thanks for this video, its really helpful, although I got worried when you said at the end 'I'm not really talking about autocorrelation'. So, if we take the contents of this video to be a description of autocorrelation, it's wrong?
Hi, thanks for your comment. Autocorrelation is an issue for prediction for three main reasons: 1. It is often indicative of the fact that there are important omitted variables (which are correlated with themselves across time) in the regression model. Omission of important variables is likely to bias your predictions. 2. Autocorrelation results in OLS estimators not being the most efficient unbiased estimators. There are other estimators - GLS for example, which will estimate the parameter values closer to their true value more of the time. Hence in order to produce the best prediction it is best to use an estimator other than OLS. 3. When building a model for prediction it is useful to be able to do proper inference on parameters, hence allowing the most robust model to be built. Inference relies on the use of standard errors most often to produce t or z stats. If there is autocorrelation in a model then the errors which statistical software programs produce by default are incorrect; often over-emphasising the statistical significance of a variable. Hence your model will be build on weak/incorrect foundations, meaning it will not be useful for prediction. Hope that helps! Best, Ben
Ben Lambert Thank you so much for your detailed answers. That is very helpful and enlightening. I realized AC is not good when it appears in building models, especially in economical model. But how about it is appears in pure statistics? I mean correlated with Spectrum as a Fourier pair? Do you have any idea about that? Thanks! seems you are an expert :-)
Hi, sorry for the very late reply. I must admit I am not sure about Fourier pairs, so am not sure I can help. My apologies for not being more helpful. Best, Ben
Could you please start subtitle your vidéos ? I'am not a native speaker and my english isn't good at all so i have a lot of difficulties and i make double effort to understand. Thank you so much for your video it's helpful.
Is autocorrelation also a problem with a cross-sectional regression? And how do you test for this? Durbin-Watson is only used for time series right? I'm running the CAPM model in SAS but don't know how to check whether this 'no autocorrelation' assumption should be tested. :)
Plus: Very informative Minus: It would have been way better, had you used actual text, instead og handwriting in (at times) barely readable text. All the best :)
![[#2024MAMA] BIGBANG (빅뱅) - 뱅뱅뱅 (BANG BANG BANG) + FANTASTIC BABY | Mnet 241123 방송](http://i.ytimg.com/vi/EQsYfiPR9uM/mqdefault.jpg)
You are amazing! Your videos have helped me so much throughout both my BA and MA degree in Economics. Now I am starting my PhD and they are still helping me a lot. Thanks so much for the work you put into it! We really appreciate it
I had to skip most of undergrad and would now be struggling through the early stages of my Master's, were it not for your videos filling the gaps! Thank you.
thank you very much .I am a french student in Maths and economics. your videos are extremely helpful for me .
Bonjour
I wanted to see a exact example done on autocorrelation function
BEN! You are the most amazing videomaker guy! Thank you sooo much!
It's very best learning please continous
Great vids, really helpful!
ugh why oh why is this stuff so complicated :(
because no one ever explains what the individual items in the equation means...such as his statement at 0:44 where he says something does not equal S...what the actual fudge is S?
@@scottyagreen The s refers to the index of Us (=error term for entity/observation s). If i =s, then the correlation equals 1 because two identitcal things are perfectly correlated. This case however is of course not relevant
.
@@scottyagreen i and s are any numbers which are not the same, they are talking about the indexes. For example: when he says the model Y_i= Alpha+ BetaX_i +U_i. Here he talks about one particular sample numbered i={1,2,3,4,5}.
Therefore U_i=error term for sample{1,2,3,4,5}
U_s=error term for sample{1,2,3,4,5} also, but as i=/u thus when i takes the value of 1, s cannot be 1, it can be anything else{2,3,4,5}.
@@scottyagreen i and s are both subscripts indicating the ith and the sth term respectively.
Example :
i=1 tests the covariance of the first term
If there are 3 terms - covariance of the 1st term will be tested against the 2nd and 3rd term
s represents the terms it is compared against so here, s= {2,3}
s shouldn't be equal to 1 as it would mean checking covariance of the 1st term (ith) with itself (sth) and that's not needed
Wish you were my lecturer, straight A's
What are the advantages and disadvantages of ACF?
Which videos are after this on autocorrelation?
Great vid as usual.. Would it be possible for you to do an illustration of the rank condition for simultaneous equations... I already know how to show if it is exactly identified or over identified using the 'order condition' However I really cant seem to understand the rank which i know is the necessary and sufficient condition. Regards.
hi, i have a question in a practice paper "What is the difference between autocorrelation and serial correlation? [2 marks]
" is the answer that there is no difference?
Hi, thanks for this video, its really helpful, although I got worried when you said at the end 'I'm not really talking about autocorrelation'. So, if we take the contents of this video to be a description of autocorrelation, it's wrong?
Thanks for your video. I would like to ask a question. So why exactly auto-correlation is a problem in prediction?
Hi, thanks for your comment. Autocorrelation is an issue for prediction for three main reasons: 1. It is often indicative of the fact that there are important omitted variables (which are correlated with themselves across time) in the regression model. Omission of important variables is likely to bias your predictions. 2. Autocorrelation results in OLS estimators not being the most efficient unbiased estimators. There are other estimators - GLS for example, which will estimate the parameter values closer to their true value more of the time. Hence in order to produce the best prediction it is best to use an estimator other than OLS.
3. When building a model for prediction it is useful to be able to do proper inference on parameters, hence allowing the most robust model to be built. Inference relies on the use of standard errors most often to produce t or z stats. If there is autocorrelation in a model then the errors which statistical software programs produce by default are incorrect; often over-emphasising the statistical significance of a variable. Hence your model will be build on weak/incorrect foundations, meaning it will not be useful for prediction. Hope that helps! Best, Ben
Ben Lambert Thank you so much for your detailed answers. That is very helpful and enlightening. I realized AC is not good when it appears in building models, especially in economical model. But how about it is appears in pure statistics? I mean correlated with Spectrum as a Fourier pair? Do you have any idea about that? Thanks! seems you are an expert :-)
Hi, sorry for the very late reply. I must admit I am not sure about Fourier pairs, so am not sure I can help. My apologies for not being more helpful. Best, Ben
+Ben Lambert Very nicely explained..Thank you very much..
Cool !!
Introduction?
Could you please start subtitle your vidéos ? I'am not a native speaker and my english isn't good at all so i have a lot of difficulties and i make double effort to understand. Thank you so much for your video it's helpful.
just use CC
Is autocorrelation also a problem with a cross-sectional regression? And how do you test for this? Durbin-Watson is only used for time series right? I'm running the CAPM model in SAS but don't know how to check whether this 'no autocorrelation' assumption should be tested. :)
nice work
Thanks!
Hi, thanks for your message and kind words. Glad to hear it was helpful! Best, Ben
Thanks!! :D
Good Video! But whats BLUE?
best linear unbiased estimator
thank you kim
Plus: Very informative
Minus: It would have been way better, had you used actual text, instead og handwriting in (at times) barely readable text.
All the best :)
Supercalifragilisticexpialidocious. thanks.
great
I can hear anything even with earphone
haha: "some other error US" 0:35
Forgot to say hi there😔
What accent is that?
british
you end not to explain nothing. You did not talk about autocorrelation
bad