Logs in series results from ARMA(1,2,1) to AR(3,2)

Asked Oct 20 '17 at 12:08

Active Dec 14 '18 at 20:42

Viewed 65 times

I am working on this dataset.

library(Quandl)
bitcoin <- Quandl("BITSTAMP/USD",type = "xts")
bitcoin.price <- bitcoin[,"last"]

When I do NOT take logs since they have increasing variance, the model is: ARMA(1,2,1) with pic:

on the other hand, if I take logs, I have:

Does this make sense? Why taking logs remove the MA component?

edited Dec 14 '18 at 20:42

phiver

asked Oct 20 '17 at 12:08

J.Ze

why don't you post this "simple series" . You night also look at http://docplayer.net/12080848-Outliers-level-shifts-and-variance-changes-in-time-series.html and https://stats.stackexchange.com/questions/18844/when-and-why-should-you-take-the-log-of-a-distribution-of-numbers – IrishStat Oct 20 '17 at 14:53
yes Sure I will do it. – J.Ze Oct 23 '17 at 10:07
1

AR and MA have competing behaviors: MA corrects slight over differentiating and AR corrects slight under differentiating. See [Identifying the numbers of AR or MA terms in an ARIMA model](https://people.duke.edu/~rnau/411arim3.htm#signatures) for details. – Firebug Oct 23 '17 at 10:24

0 Answers0