RU-vid recently changed the way my content will be monetised. My channel now needs 1,000 subscribers. So it would be amazing if you show your support by both watching my videos and subscribing to my channel if you haven’t done so already. Monetising my videos allows me to invest back into the channel with some new equipment so this small gesture from you will be extremely huge for me. Many thanks for your support….CrunchEconometrix loves to teach, support my Channel with your subscription and sharing my videos with your cohorts.
@@CrunchEconometrix Thanks Professor for your kind reply. Could you please explain to me why do you put d(gdp) as a dependent variable even though you were working gdp data set as it was seen in the identification video? Anyway, I am trying to forecast an autoregressive model of inflation from the monthly frequency data set. I obtained two very large different results using the same order but putting one without d and another one d on the dependent variable in the equation.
Please, I have some questions. Can I run ARIMA model for panel data? I have GDP data for several provinces for a period of 10 years and I want to predict possible future GDP for each province
@@CrunchEconometrix Secondly, the ACF and PACF plot for some province doesn't have any spike outside the error bounce. What does this mean, and how do I fit model for this kind of scenario? Thanks.
firstly ı m really appriciate you because of your incredible efforts, my question ise significant coefficients, how can I decide or see them how many they are, please answer, thank you ...
Hi YDS, from the p-values or t-stats of the coefficients, the significant coefficients can be easily seen. I prefer using the p-values either at 1% (very strong significance), 5% (strong significance) or 10% (weak significance). Also, follow my explanations for better understanding..may I know from where (location) you are reaching me?
Good to hear this feedback, Ope. Deeply appreciated. Please tell others to subscribe by sharing my videos with your friends and academic community on social media...gracias! 💕 😊
Good to hear this feedback, Ope. Deeply appreciated. Please tell others to subscribe by sharing my videos with your friends and academic community on social media...gracias! 💕 😊
mam! I'm in just love with your vedios. it's so much easy to understand, no boredom no headache, just one click on ur vedios and going through it made the hardest Econometrics to the easiest one for me (from Bangladesh) Thankyou💜
Thanks, Sidrat for the encouraging feedback. My aim is always to simplify the understanding of applied econometrics. Glad my approach is yielding some positive outcomes. Love from Nigeria 💗
Thanks Gift, for the positive feedback...deeply grateful. Kindly share with your students and academic colleagues. May I know from where you are reaching me from?
I’m confused when i choosed Arima(1,1,1) Arima(2,2,1 Arima(1,2,1) Arima(1,2,2) So i get Ar(1) or Ar(2) Alse Ma(1) and Ma(2) Those are (p and q) how i runout d means difference or i. I confused "The reviews, they can only be made in Ar and Ma. I don't see no other place they gets produced. “Difference(d)
Hi Mayank, thanks for the encouraging feedback. Deeply appreciated! I'm a lady, though (lol). Please may I know from where (location) you are reaching me?
@@mayankshukla205 Awesome! I'll appreciate it if you can share the link to my RU-vid Channel with your students and academic community in India 🇮🇳...May God bless you as you do, amen 🙏
hi mam, I want to ask if on the 1 difference correlogram, none of the AR (partial autocorrelation) and MA (autocorrelation) cross the line, how will we determine the model to be used for forecasting? and on the correlogram, does the probability have to be significant less than 0.05? thank you so much for helping me🙏🏻
Dea, the guidelines for model selection are quite clear. Kindly follow the steps, or you may want to check out other online resources for more information. Thanks.
Thanks Pralhad, for the positive feedback and remarks on my video. Deeply appreciated! Unfortunately I have no idea about SARIMA. Please may I know from where (location) you are reaching me?
Thanks for the clear video, it helped me a lot for my work. However, I'm facing a situation not presented here and I would really appreciate your insights. If I'm comparing two SARIMA models that estimated a different number of coefficients, on what criteria should I base my selection of the best model? Thanks
@@CrunchEconometrix i have a small doubt... Forecasting means we predict a future value.. That is if our data is till 2007, we can predict the value for 2008 but through ARIMA forecasting.. We didn't do this.. Why?
hi! thank you so much for your videos. If i want to test arma models instead of arima, the steps and clicks are the same? what does the c in the command for the first tests presented means?
I have a question. I have data on CPI which follows a similar trend with GDP. When working with the difference of CPI I observe the same exponential decay within ACF and PACF. However, as soon as I take the log of CPI and then work with the difference of this log, the ACF does not decline as quickly, always significant between lags 1 and 11 then drops down. PACF, however, declines more rapidly. Many economic/financial time series grow approximately exponentially, so taking the log makes sense. In our examples covered in classes, we work with GDP. We have always taken the log of GDP due to the reason mentioned previously, and we have been told CPI grows exactly the same (exponentially). So, in order to perform the ARIMA Box-Jenkins approach, do I proceed with first difference CPI (which seems to fit the trend better from correlogram as expected) or proceed with first difference log CPI (which is a transformation suggested for economic time series that grow approximately exponentially)? Many thanks.
CrunchEconometrix I understand now. So I could argue that despite the fact in my case CPI may have a better “fitting” Correlogram than Log CPI, I decided to continue going with Log CPI for own reasons (such as attempting to stabilise the level and the variance of the series of CPI) even if the rest of the diagnostics (LM test, inverted AR & MA analysis, ARCH effects etc) show similar conditions for both CPI and log CPI? Because as you mention, it is an art more than a science, so it would not be a “mistake” to consider log CPI despite its ACF and PACF properties. Also many thanks for your quick responses, much appreciated.
Thank you for your videos...great content. I am wondering two things: 1. do you have videos on how to tweak this approach for SARIMA? 2. what software do you use? Thank you again!
Thank you for the video, it's beneficial :) I want to know why you did not select the ARIMA LAGS automatically by processing automatic ARIMA FORECAST? THANK YOU
Houria Mohamed Hi girl, thanks for the kind words of encouragement. Not choosing automatic lags is just a matter of choice. I didn't want to do any automatic forecast. Otherwise, I would have done so...and thanks for your subscription too!💕👍🏽
Thank you for your prompt answer :) actually i asked this question because I am writing my thesis about stock exchange, after seeing your videos I got confused :( I have applied the automatic ARIMA after differencing the series and I got ARIMA(4,1,4) In this case i do not need to apply the correlogram test for the residual ?
Houria Mohamed Don't be confused. Automatic ARIMA modeling is still in order. It's just that I don't use it. But, go ahead and test the correlogram of the residual to ensure that it's flat.
Madam, I have a question. If continuously autocorrelation or serial correlation arises after conducting Breusch-Godfrey Serial Correlation LM Test, then what I will do? I have seen Obs*R-squared 1948.904 Prob. Chi-Square(2) 0.0000 and R-squared 0.857792, F-statistic 1239.293 Prob(F-statistic) 0.000000 DW Stat. 1.56 I have calculated Heteroskedasticity Test also. This also Zero. Obs*R-squared 782.5454 Prob. Chi-Square(9) 0.0000. It indicates Heteroskedasticity problem. I have already applied all the time series data after converting into natural logarithm. I did not apply Panel data methodology. Panel data methodology, now learning from your video. Thank You
@@CrunchEconometrix Multiple Regression Analysis on 4 Macroeconomic variables. After that I had tried to find out the residuals for autocorrelation and multicollinearity based on LM test. I have tried to find out heteroskedasticity test also. But I can not solve the multicollinearity problem.
Hi ma’am, so if I am using the daily data, what would be the number of lags ? And thank you so much for your effort, it’s very clear and understandable :)
Hi Divaa, thanks for the encouraging words and feedback on my videos. Deeply appreciated! Kindly watch my video on "Optimal Lag Selection" for more insights. Thanks.
This video is well explained from the estimation of the 4 models to the summary extracted to ppt. Significant coefficients are obtained from each regression. You may want to watch the clip again and jot down some notes.
@@CrunchEconometrix Oh so nice to get a response from you! yes i have watched it and i can't thank you enough, but the problem is i can select only one model , only one spike in PACF. But i gotta compare two models in theory, so can i instead use “Proc>Automatic ARIMA Forecasting, and compare my model and the model that eviews suggested me?
Hai ma'am I wanna ask, if the correlogram is used 2nd difference, if we want to estimate the equation, it is written "D(GDP) or D(GDP,2)" thank you so much ma'am, may you always be healthy
Hello and thanks for the video, really helpful. I currently working on my master thesis. The thesis is about exchange rate. Basically, i'm trying compare OLS forcasting and ARIMA forecasting. After analysing my time serie, i found out 2 order differencing is needed and my arima model is (0,2,0). So i was wondering is this is possible and what does that mean? Also I use the auto.arima forecasting in eviews for my serie and i got complety different results for my analysis. The auto arima is using 1 order differencing. But i checked by using the ADF test and the first order differencing serie is not stationnary. Finally, i used auto.arima model in R and I got a complety different result for Eviews. And still R is also under-differencing the serie if i trusted my ADF test. Can i trust the auto.arima? if yes, R auto.arima or Eviews auto arima? I can't wait to read you. Many many thanks in advance Fred
Fred, you should have watched the prerequisite video as advised. Please watch it to understand the basics of ARIMA modeling because you'd have known that ARIMA (0,2,0) implies that the series has 0 AR and MA processes. Hence, the ARIMA technique is not applicable.
ok ok thanks!! I have another question. How can i deal with that? I mean the serie in difference twice to get it stationnary, but there is no lag out of boundary for the second differenced serie?. Need to differenciate more ?
Mam, i am having doubt over choosing one of two models as, my ARMA (1,1) is having lower adjusted R sq. value and ARMA(2,1) is having lower sigma sq. value and both have AIC and SBIC value same upto 3 decimal places and while considering the 4th decimal place of AIC and SBIC of both models ARMA(2,1) becomes lower. what should i choose ? pls help.
Hi Ajit, I have given the guide to choosing the most appropriate ARIMA model. There will be some differences between both models (however little) so, kindly watch the video again and decide.
Can you tell me how to find ARIMA(0,2,1)??? as the d - variable is differenced twice how to do it? one more doubt.. how to find the most significant coefficient? m not able to understand the explanation
Hi Ramya, generate the 2nd difference of that series and estimate this equation: d2Y c ma(1). The AR component is dropped because it is 0. My explanation on the "most significant coefficients" is very clear. Select the model that has the highest number of significant coefficients. I'll suggest you re-watch the clip while taking some notes for better understanding. Thanks.
@@CrunchEconometrix thank u so much mam. though i listen clearly to ur audio for significant coefficient i am not able to find. can u plzz help me. m able to see coefficient and probability values. i want to know how u r getting 0 and 2 for significant coefficient. how to find
Buvanesh Chandrasekaran Uwc Buv... Please tell others about my RU-vid channel by sharing my videos and link. I'll really appreciate that in reaching out to a wider audience😃
@@CrunchEconometrix I made effort and got the result..but in that am not getting the value for sigmasq rest I got..can u give me ur mail id so that I can send my data along with calculations and clarify my doubt
@@ramyaramamurthy7756 This shows that you are yet to understand the basics of ARIMA. Watch that video and take notes while doing so. You can't get a volatility coefficient since AR and MA are 0.
Yes, link is disconnected since August 2019 due to unethical behaviour of persons who attempted hacking my Google drive on several occasions. You can access data on my website. While some are free to download, some are on paid download. Here's the link crunchecometrix.com.ng/shop/
@@CrunchEconometrix i want to do arima on same data set for educational and practice purpose.. could u plz give it to me... plz.. it will really hlpful for clearing my doubt... plz my mail id is surajit.m.com@gmail.com