Time series in Stata®, part 5: Introduction to ARMA/ARIMA models

[@marinegattoni8781]

Now I finished watching the video and I have to thank you again, you may have just saved my assignment's grade. Best video on youtube EVER!

[@erickwhite4462]

Hi guys. Great video! Excuse me for the stupid question, but how do I interpret the coefficients at 7:28? For instance, what do the AR lag 1 (0.804) and the MA lag 2 (-0.179) entail? Thank you very much! ;) Erick

[@meshackamimo1945]

Than,Sir. Am happy you have made strata sound quite simple to follow when going aroma. Kindly do a similar, simpler example on how to do kalman filters in strata. I will really appreciate. God bless.

[@marinegattoni8781]

Thank you, big fan of your work, love stata so much thanks for sharing my passion

[@ericobarrett]

Because the series has one unit root, the differentiation/integration order should be 1. The best fit, according to AIC and BIC (estat ic), is an ARIMA (1,1,1) with no constant. "arima D1.ln_wpi, ar(1) ma(1) nolog noconstant"

[@ricardoortega3266]

thank you for the effort, is there AR(p) and AFIRMA tutorials in the channel?

[@wanjadouglas3058]

This was amazing.

[@sutanbanuara9919]

Thanks for the tutorial. what is the criteria to choose the best ARIMA model (2,1,2) or (3,1,3) or else? is it the log likelyhood? higher/lower? the AIC? the BIC? higher/lower?

[@dxluva]

I am having trouble transforming my respective variables into the Natural Log form. I used the gen ln_x=ln(x) on simplier data sets and it worked just fine, however, it is now telling me the variable is already defined. Any Suggestions?

[@guiriamine1536]

thank you for your explanation, but I have a question please .i want to model my time series ( stock returns) with ARMA-GARCH. IN residual analysis, I found some p-values for the Q test (Ljung-Box), are significant and others not, should I consider or reject H0? It's normal ??? I continue my work ?? I must differentiate the series? to pass to the GARCH modeling. THANK YOU

[@JamesRivington1]

Very informative. What would you suggest to do if after first differencing the correlogram and partial correlogram have no significant terms?

[@Mizzybelly]

Great! Very helpful, only I have more of a parabolic line instead of a exponential, how to transfer/deal with that?

[@jesfjo]

I'm thinking of perhaps some sort of combined graph, that combines line graphs for each person and makes the x-scale look coherent in some way. I just need a visualization of the dataset. Thanks for any input.

[@nataliaancaovaldes4052]

Good afternoon. I need to know wich is the commands or syntax iteration. I have to do a homework: process ar (p)n =500, sigma 3 and 5000 iteration , Please your help !! u.u Thank you very much

[@mihoda]

Great stuff, but I think you meant to run the arima on ln_wpi rather than wpi.

[@cohenjota]

I belive... the same... whos right?

[@prabrai8837]

Hi - thank you for your video, it was very informative. I have 2 questions regarding the PACF. - would you ever use lag 4? even though lag 3 is not significant? - also, if lag 4 was significant but had a value of -0.4, would you still use it? (with lag 3 again not being significant? Thanks again! :)

[@statacorp]

+P Rai Would you ever use lag 4 even though lag 3 is not significant? YES. Think for example of quarterly data for sales. The third quarter might be insignificant but the last quarter includes data for the holidays which are important. If lag 4 was significant but had a value of -0.4, would you still use it? (with lag 3 again not being significant? This one is more tricky. You should not include variables in your model based on their significance. You should include them based on theory or some other criteria. For instance, the predictive power of your model might be greater with the insignificant variable than without it. The criteria used depends on the problem at hand.

[@reneordosgoitia]

Hi everybody, i am trying to do an arima model, with the price of a stock. When i made the AC and PAC test; in order to obtain AR and MA, it doesn´t provide any results out of the coinfidence level of 95%. What can I do? :) I am new in the econometric world

[@niemaelamin74]

+Rene Ordosgoitia maybe you overdifferenced your series check this out people.duke.edu/~rnau/411arim3.htm

[@cindyli7806]

You are amazing

[@jhsr2501]

Thank you sir for this useful video. I have a question: How to determine Extended autocorrelation in ARMA model using stata? In Tsya (1984), it was mentioned that ACF may not be useful for determining the order of ARMA. It will be really useful if you can come up with a video on determining the order of ARMA model

[@chavaliramamurthy512]

hi...good tutorial ....why should you go for D.Ln(wpi) in order to make the data stationary? Why not directly go D.wpi bcos log only changes the scale, not get you stationarity. whereas, differencing is wat brings you stationarity...moreover, how are you gonna explain difference logged wpi in analysis while d.wpi ..is simply the quarterly change in the wpi index :)

[@statacorp]

+Chavali Ramamurthy Thank you for the feedback. The Wholesale Price Index is sometimes modelled on a log scale and the decision to use the log scale was made somewhat arbitrarily. Many people model d.wpi and we agree with your comment that it would be easier to interpret d.wpi than d.ln_wpi.

[@guiriamine1536]

+Chavali Ramamurthy thanks sir for your explanation . i HAVE a question after estimating arma , how should i analyse the residuals . Qs squared residuals on stata , thanks ??

[@JMRG2992]

what is sigma in this regression?

[@jesfjo]

Could you point me in the right direction when trying to graph data with multiple measurement for several people, measured at different times. Such as this: person, time, value 1, 10:10, 2.5 1, 10:30, 3.1 2, 09:30, 3.2 2, 11:20, 2.7

[@zoozolplexOne]

good !!!

[@meisterkavar]

I have to say that this video is totally useless, because without actually telling us how interpret the figures in the end of the video (which is to me the most and furthermore only important issue which this video should focus on) this manual does not help at all......

[@slts94]

a better model would be ARIMA(2,1,0)