Ritvik, you really have a gift for teaching complex topics in such simple terms. Seriously, I'd been trying to find an understandable lesson, and yours was godsent! Thank you very much for taking the time to help us!
I'm doing research and it's involve with some of the concepts you mentioned, I've never been felt how easy to understand these concepts till I saw your video!! Big Thanks to you ,, please keep posting more videos for the sack of science research and education.
Oh my Lord!!!! This is amazing! They could pay people money from here to the moon and they wouldn't be able to explain this concept so concisely. Best explanation of AR Model I've heard. Thank you so so much!!
Most error in prediction models answers only how many % chance an event happen. BUT THEY NEVER ANSWER YOU the magnitude WHAT IF THE SMALL CHANCE HAPPEN. Some events like 2020 here rarely happened, but when breaking out, its magnitude swipe out everything. HAHA
ua-cam.com/video/nnwqtZiYMxQ/v-deo.html . Case study on Amul during covid. Every hard hit comes with momentum that can destroy us or push hard to be the best of all time.
I’m a data scientist who worked through the pandemic in a critical infrastructure industry. On the other side now, can confirm, standard methods rendered results like 1+1=purple.
for the AR model you made for m(t), would this be an AR(4) model because there are 4 lags, or would it be an AR(12) model because the largest lag is 12 periods before the current time t?
I think in this case, the model would be considered an AR(12) model. Even though there are only 4 significant lags (1, 2, 3, and 12), the largest lag is 12 periods before the current time t. When specifying an autoregressive model, the order of the model is determined by the maximum lag included in the model, which in this case is 12. The AR(12) model would include all lags up to the 12th lag, with some coefficients possibly being zero or near-zero for the insignificant lags.
@@phut7755I would beg to differ. We denote an autoregressive model as AR(p), where p denotes the amount of lagged variables included in the model, which in the case of the example from this video is 4. Hence it is an AR(4) model.
Great video! Just one thing I didn't completely understand. when trying to find the model of Mt, where do the beta values come from? Thanks! (timestamp: 7:18)
Hi Ritvik, thank you for these viedos. It seems like this one should be the third one in the time series playlist, after ACF and PACF are introduced, but before the coding demo which already references AR.
Amazing easy explanation my friend! It's a pity that you didn't explain the beta coefficients in detail, but I understood the concept very well :-) Thank you for your help.
Hey Ritvik! I had a doubt, what is the difference between a simple exponential smoothing and an AR model? Simple exponential smoothing predicts the next value as a linear function of the previous values, but weighted. AR Model also predicts the next value as a function of the previous ones. So is exponential smoothing a subset of AR model or how does it go?
In exponential smoothing, the used weights follow an exponential model. In AR, by contrast, there's no constraint on these weights. So as you suggest, exponential smoothing in this context could be a special case of AR.
Hi, great videos! I am following the series and one thing that is not clear is that this milk chart seems to have a seasonality. My question is, if you can model it with just an AR model why do I need the "s"arima model? I will answer my own question, I think I understood. The SARIMA is just applying "AR" "I" and "MA" over the seasonal lag. So for example if I have an yearly 12months seasonal data using just AR(12) would calculate the regression over all steps/months 1,2,3,4,..12 but if I have S"AR"(12) it will just calculate the regression on the 12th lag
amazingly simple explanation, thanks! My trouble so far is understanding what the beta coefficient(0) or intercept is. can you explain it briefly please?
Hi sir, seeking for clarification here, why is it that AR Models can only be applied to stationary time series? This one here isn't stationary due to seasonality, but it seams like the seasonality helps in the prediction, due to the 12th month adding an additional month that helps predict the current month?
Thank you for the video. From the video, I have two questions in mind, 1. Is AR model built from PACF? 2. Can we also build AR model from ACF? Hope to hear some from you!
Later videos say that AR cannot be used on a seasonal model which this clearly is. But the model is based on the seasonality. So can it be used or not?
I learned the info from your last video that to apply a time series, the data shouldn't be seasonal. Isn't the data here somewhat seasonal to apply a time series forecasting model ?
Thank you so much for your video - I am actually watching your whole TS playlist and it helps me so much!! I have just one little question regarding the model you presented us with at the end: Shouldn't it be minus ß2 and minus ß4 as mt-2 and mt-4 have a negative direct influence on mt, which is then expressed in their coefficients? Would be great if you or anybody else could help me out. Thanks! :)
I'm having a problem with the definition of order of AR, MA and ARMA time series forecasting processes. Imagine we have a time series with data from January to December, and we're in July, trying to predict August. When we say AR(2), are we using lags relating to July and June, or can those two months be any month between January and June?
If AR model can only be applied on stationary data set, how come the example used in this video is clearly non-stationary? The dataset example has yearly seasonality, correct?
Just a question, you said in a previous video that if we want you use AR or MA model, the serie has to be stationnary. Here it's obviously not so why we use a AR model please ?
Great Video! My questions are: 1) In your first video about ACF and PACF, as long as there is a time series, i could plot ACF and PACF regardless on whether its stationary or not by my understanding. In this episode, the time series need to be stationary in order to implement AR model. Why is that? 2) In my case to analyze stock price, the first step is to plot ACF and PACF. Do I need to make stock pice stationary in order to perform ACF and PACF? Thank you !
I maybe wrong but i think he was just checking the time series data for stationarity. Becuz if its stationary we go for OLS and if not stationary we try and apply ARDL model to the time series data.
What an amazing explanation sir.. Great sir.. Sir plz make video on cointegration especially Johensen cointegration.... What is difference between VAR AND AR.. PLZZZZ HOPE TO SEE YOUR REPLY
@ritvikmath In the video for stationarity, you mentioned that we need stationarity to apply AR/MA models to the time series. Furthermore, for a time series to be stationary, it had to have the following criteria : 1. The mean / expected value remained constant 2. The variance remained constant 3. There was no seasonality Yet in this video, we are using an AR model to solve a problem which is completely seasonal. This felt contradictory to your stationarity video.
But aren't you supposed to stop at the first insignificant lag, in this example 2 lags were significant then lag 3 was not so a good model should be AR(2) and not AR(4) right ?
It would seem to me that from your discussion of the use of the PACF to identify the important contributors that you have missed the lags at t-24 and at t-36 unless your analysis makes the assumption that the quantity has a periodicity of one year. But you didn't discuss periodicity in your approach to the PACF.
Its for the first time that I have seen someone explaining econometrics in such a simple but yet in a comprehensive manner. You are a life saver.
Ritvik, you really have a gift for teaching complex topics in such simple terms. Seriously, I'd been trying to find an understandable lesson, and yours was godsent! Thank you very much for taking the time to help us!
I am absolutely amazed. Thank you so much for this
I'm doing research and it's involve with some of the concepts you mentioned, I've never been felt how easy to understand these concepts till I saw your video!! Big Thanks to you ,, please keep posting more videos for the sack of science research and education.
is your research by any chance is on ARx model? doing the same :p
you’re a lifesaver!!! the amount of light bulb moments I have in your videos is insane
So well explained again - you are brilliant at explaining the concepts in a way that's easy to understand - THANK YOU!
Glad it was helpful!
It is incredible how well you teach. These videos are fantastic, thank you
Glad you like them!
Gem of a series for anyone studying about time series!!
this is the easiest but best video I saw to understand AR Model! thank you very very much!
Glad it helped!
Bro, this was easily the best explanation I've ever heard so far. Thanks a lot!
Oh my Lord!!!! This is amazing! They could pay people money from here to the moon and they wouldn't be able to explain this concept so concisely. Best explanation of AR Model I've heard. Thank you so so much!!
Brilliant explanation. So easily explained this confusing topic.
It's amazingly simple and clear explanation of such a elusive topic! Thank you very much
Really a gentle but a very powerful and intriguing intro to the AR model. Thank you.
Thankyou so much, This video was of great help. one of the best material explaining time series forecasting. :)
Taking your videos help in 2023🎉❤thak you ritvik or ritik sir
This is so helpful!! You cleared all my doubts. Thank you very much for making this.
Glad it was helpful!
came here for copper, found gold instead. You doing a great job with these video my friend. thanks
Wow! You are a principality, with due respect this is mind blowing
2020 hit us so hard no statistical model could hold. I bet even the milk demand is a total mess now!
Most error in prediction models answers only how many % chance an event happen. BUT THEY NEVER ANSWER YOU the magnitude WHAT IF THE SMALL CHANCE HAPPEN. Some events like 2020 here rarely happened, but when breaking out, its magnitude swipe out everything. HAHA
Although some model may not hold, this will help us factoring in the effects of such events when we deduce other similar models.
@@anthonyng3705 That's what you call Excpected Shortfall in finance. Expected loss given a tail event
ua-cam.com/video/nnwqtZiYMxQ/v-deo.html . Case study on Amul during covid. Every hard hit comes with momentum that can destroy us or push hard to be the best of all time.
I’m a data scientist who worked through the pandemic in a critical infrastructure industry. On the other side now, can confirm, standard methods rendered results like 1+1=purple.
Thank you, very nice explanation.
Q: How do you draw the "error" lines (red dotted) in the ACF plot? What is this threshold for significance?
Thank you for this series ! ❤️❤️❤️
You are so welcome!
Thank you so much for your clear and well put together videos
Not a problem :)
Thank you so much 😊
This video is amazing. Thankyou for explaining this so well
for the AR model you made for m(t), would this be an AR(4) model because there are 4 lags, or would it be an AR(12) model because the largest lag is 12 periods before the current time t?
I think in this case, the model would be considered an AR(12) model. Even though there are only 4 significant lags (1, 2, 3, and 12), the largest lag is 12 periods before the current time t. When specifying an autoregressive model, the order of the model is determined by the maximum lag included in the model, which in this case is 12. The AR(12) model would include all lags up to the 12th lag, with some coefficients possibly being zero or near-zero for the insignificant lags.
@@phut7755I would beg to differ. We denote an autoregressive model as AR(p), where p denotes the amount of lagged variables included in the model, which in the case of the example from this video is 4. Hence it is an AR(4) model.
Great video! Just one thing I didn't completely understand.
when trying to find the model of Mt, where do the beta values come from? Thanks! (timestamp: 7:18)
Great video man ! Big love from Saudi
Thanks a lot. You're undoubtedly a genius.
So great sir, hope to see more video about time series from you, it is really benefits for me
You made my intuition clear. Thank you
Very nice explanation. Thank you a lot!
Hi Ritvik, thank you for these viedos. It seems like this one should be the third one in the time series playlist, after ACF and PACF are introduced, but before the coding demo which already references AR.
Very good, well explained.
Glad it was helpful!
Hi! The milk graph shows seasonality. I'm wondering how could you use AR model on a nonstationary time series. Thank you.
I have the same question
That's what ARIMA model is for. He has a video on that.
this stationary time series the mean is fairly constant
Hello. If there is seanality you could just do a second difference to remove it.
well. correct me if im wrong. i dont think AR model can skip lags tho, meaning it needs to start from t-1 and follows in time order i believe
You are a great teacher
Thank you so much, brilliant!!
Really such a wonderful and understandable vedio this is.
Excellent video!
Amazing easy explanation my friend! It's a pity that you didn't explain the beta coefficients in detail, but I understood the concept very well :-) Thank you for your help.
Wonderful explanation!!!!!! do you have video explaining the differences between AR-MA-ARMA-ARIMA?
before talking about AR model, the time series must be STATIONARY !
AR and MA models are based on stationary time series
Holy man, you are a natural!!! Thanks a lot!!!!
Hey Ritvik!
I had a doubt, what is the difference between a simple exponential smoothing and an AR model?
Simple exponential smoothing predicts the next value as a linear function of the previous values, but weighted. AR Model also predicts the next value as a function of the previous ones. So is exponential smoothing a subset of AR model or how does it go?
In exponential smoothing, the used weights follow an exponential model. In AR, by contrast, there's no constraint on these weights. So as you suggest, exponential smoothing in this context could be a special case of AR.
Very useful. Thank you!
Great video! Thank you very much!
Brilliantly explained
Great video, keep going.
Great explanation! Thank you very much!
Well explained. Thank you very much you may have saved my assignment haha
Hi, great videos! I am following the series and one thing that is not clear is that this milk chart seems to have a seasonality. My question is, if you can model it with just an AR model why do I need the "s"arima model?
I will answer my own question, I think I understood. The SARIMA is just applying "AR" "I" and "MA" over the seasonal lag. So for example if I have an yearly 12months seasonal data using just AR(12) would calculate the regression over all steps/months 1,2,3,4,..12 but if I have S"AR"(12) it will just calculate the regression on the 12th lag
Thank you very much! it is a very well explained and useful video!
Thanks for this very clear explanation!!!
great video as always
amazingly simple explanation, thanks!
My trouble so far is understanding what the beta coefficient(0) or intercept is. can you explain it briefly please?
Excellent video. Clearly explained and loved the crayola markers.
For this, would you use Level data or first differences?
Thank you
thanks a lot for your work
You are welcome!
great video!
Thanks!
really very helpful
Glad you think so!
Great explanation
Hi sir, seeking for clarification here, why is it that AR Models can only be applied to stationary time series? This one here isn't stationary due to seasonality, but it seams like the seasonality helps in the prediction, due to the 12th month adding an additional month that helps predict the current month?
Thanks this is so informative!
In this example the data is seasonal, does this mean we need to make the data stationary before we use the PACF plot?
Thank you for the video. From the video, I have two questions in mind,
1. Is AR model built from PACF?
2. Can we also build AR model from ACF?
Hope to hear some from you!
AR model is identified or built by PACF plot
And MA model is identified or built by ACF plot...
Always remember
Thanks for the lesson. Help me a lot. ;)
The PACF appears similar to Tornado plot in uncertainty analysis.
great job sir!
Later videos say that AR cannot be used on a seasonal model which this clearly is. But the model is based on the seasonality. So can it be used or not?
thanks a lot, sir! helped me a lot, to understand concept
This is amazing, thank you.
Superb
Hello sir, Won't the t-2, t-4 terms get negative sign, as they are in the negative direction?
The coefficient can be a negative value (e.g. b2 = -0.6). No need to use negative signs
Very good video!!
For this AR model what will be the p value? That is, AR(p) -> AR(4)? Is that correct?
How do we estimate the variance of the white noise from the given data?
This helped me a lot. Do you have any recommended bibliography?
I learned the info from your last video that to apply a time series, the data shouldn't be seasonal. Isn't the data here somewhat seasonal to apply a time series forecasting model ?
Thank you so much for your video - I am actually watching your whole TS playlist and it helps me so much!! I have just one little question regarding the model you presented us with at the end: Shouldn't it be minus ß2 and minus ß4 as mt-2 and mt-4 have a negative direct influence on mt, which is then expressed in their coefficients? Would be great if you or anybody else could help me out. Thanks! :)
i guess that the beta coefficients may be negative
I'm having a problem with the definition of order of AR, MA and ARMA time series forecasting processes. Imagine we have a time series with data from January to December, and we're in July, trying to predict August. When we say AR(2), are we using lags relating to July and June, or can those two months be any month between January and June?
yes, Video is superb. How can we select order of AR model from PACF and same for MA model from ACF.
please make more time series video! It really helps! and there is no much time series video out there at all
me also like much time series video. Hope make more video for knowledge.
Seems like AR is for capturing seasonality.
If AR model can only be applied on stationary data set, how come the example used in this video is clearly non-stationary? The dataset example has yearly seasonality, correct?
A nice introduction. Maybe you could use the example data and show the prediction curve to get a sense of the outcome.
this is so nice if you try to learn math without confusion
Just a question, you said in a previous video that if we want you use AR or MA model, the serie has to be stationnary. Here it's obviously not so why we use a AR model please ?
Great Video! My questions are:
1) In your first video about ACF and PACF, as long as there is a time series, i could plot ACF and PACF regardless on whether its stationary or not by my understanding. In this episode, the time series need to be stationary in order to implement AR model. Why is that?
2) In my case to analyze stock price, the first step is to plot ACF and PACF. Do I need to make stock pice stationary in order to perform ACF and PACF?
Thank you !
I maybe wrong but i think he was just checking the time series data for stationarity. Becuz if its stationary we go for OLS and if not stationary we try and apply ARDL model to the time series data.
Hi, so it is okay to select 1,2,4,12 ignoring the lags in between?
What an amazing explanation sir.. Great sir.. Sir plz make video on cointegration especially Johensen cointegration....
What is difference between VAR AND AR.. PLZZZZ HOPE TO SEE YOUR REPLY
@ritvikmath In the video for stationarity, you mentioned that we need stationarity to apply AR/MA models to the time series. Furthermore, for a time series to be stationary, it had to have the following criteria :
1. The mean / expected value remained constant
2. The variance remained constant
3. There was no seasonality
Yet in this video, we are using an AR model to solve a problem which is completely seasonal. This felt contradictory to your stationarity video.
8:29 where has the mean term gone? Looks like it is represented by beta sub 0.
do i need to find the model again every time a new data is added or can i use the model to predict the value for the next couple period?
But aren't you supposed to stop at the first insignificant lag, in this example 2 lags were significant then lag 3 was not so a good model should be AR(2) and not AR(4) right ?
How do you calculate the red bands, so that you can check which lagged value has an impact on the model?
thx for answer :)
Much appreciated :-)
Is this a AR(12)?
It would seem to me that from your discussion of the use of the PACF to identify the important contributors that you have missed the lags at t-24 and at t-36 unless your analysis makes the assumption that the quantity has a periodicity of one year. But you didn't discuss periodicity in your approach to the PACF.
can we predict commodity prices based on weather?
I really liked the video, maybe next time you could finish the example with some actual numbers