cipherap15
New member
- Jun 18, 2026
- 0
- 0
My last post was long and messy and even confused me. This is where I’m at if anybody can help.
1. For AR(1) model if Autocorrelation is sigificantly different then 0, then serial correlation exists and we try AR(2) and so forth. (Mentioned throughout chapter). When all the autocorrelation do not signficantly differ then 0 then that AR model fits.
2. Focusing on Random walk with drift case for an AR(1) model. Examination of autocorrelation is a well known prescription for inferring whetehre or not a time series is stationary. Typically for Stationary, Either Autocorrelations at all lags are indistinguishable from 0 OR the autocorrelations rapidly drop off to 0 as the lags increase. (Page 430 - 5.2 - Unit root discussion)
3. If returns had come an AR(1) time series, the FIRST autocorrelation would have differed significantly from 0 and the autocorrelations would have declined gradually. (Pg 438 - End of example 14)
4. For MA(1) model all autocorrelations will be 0 except the first one. MA(0) all autocorrelations will be 0. MA(2) all autocorrelations will be 0 except the first two. (Page 436 - Section 6.2)
Number 3 is the one that confuses me and just seems wrong knowing all the information. If the first autocorrealtion was significantly different then 0 as mentioned in the paragraph, there would be serial correlation and it could NOT be an AR(1) model. The question states this is how you can tell the difference between an AR(1) model and if it’s a MA(0) model. I think you cant tell the difference between the two. They both should have all autocorrelations being 0.
I was a little confused about two as well when it mentioned the first few lags can differ from zero then it rapidly drop off. But I figure this is only the case for AR with Drift.
Anybody?
1. For AR(1) model if Autocorrelation is sigificantly different then 0, then serial correlation exists and we try AR(2) and so forth. (Mentioned throughout chapter). When all the autocorrelation do not signficantly differ then 0 then that AR model fits.
2. Focusing on Random walk with drift case for an AR(1) model. Examination of autocorrelation is a well known prescription for inferring whetehre or not a time series is stationary. Typically for Stationary, Either Autocorrelations at all lags are indistinguishable from 0 OR the autocorrelations rapidly drop off to 0 as the lags increase. (Page 430 - 5.2 - Unit root discussion)
3. If returns had come an AR(1) time series, the FIRST autocorrelation would have differed significantly from 0 and the autocorrelations would have declined gradually. (Pg 438 - End of example 14)
4. For MA(1) model all autocorrelations will be 0 except the first one. MA(0) all autocorrelations will be 0. MA(2) all autocorrelations will be 0 except the first two. (Page 436 - Section 6.2)
Number 3 is the one that confuses me and just seems wrong knowing all the information. If the first autocorrealtion was significantly different then 0 as mentioned in the paragraph, there would be serial correlation and it could NOT be an AR(1) model. The question states this is how you can tell the difference between an AR(1) model and if it’s a MA(0) model. I think you cant tell the difference between the two. They both should have all autocorrelations being 0.
I was a little confused about two as well when it mentioned the first few lags can differ from zero then it rapidly drop off. But I figure this is only the case for AR with Drift.
Anybody?