Could someone tell me what the term 'persistence' mean in time series analysis? It's regarding econometrics and applied regression.
2 Answers
Roughly speaking, the term persistence in time series context is often related to the notion of memory properties of time series. To put it another way, you have a persistent time series process if the effect of infinitesimally (very) small shock will be influencing the future predictions of your time series for a very long time. Thus the longer the time of influence the longer is the memory and the extremely persistence. You may consider an integrated process I(1) as an example of highly persistent process (information that comes from the shocks never dies out). Though fractionally integrated (ARFIMA) processes would be more interesting examples of persistent processes. Probably it would be useful to read about Measuring Conditional Persistence in Time Series in G.Kapetanios article.
-
3$\begingroup$ In graduate level textbooks the persistence usually is the synonym of unit root. $\endgroup$– mpiktasCommented Jun 15, 2011 at 10:57
-
$\begingroup$ The link to the article seems to be down. can you pls update $\endgroup$– SandeepCommented Jan 22, 2019 at 14:45
-
$\begingroup$ I would appreciate if you could answer this question: stats.stackexchange.com/questions/388968/… Thanks @DmitrijCelov $\endgroup$– ebrahimiCommented Jan 25, 2019 at 15:17
A persistent series is one where the value of the variable at a certain date is closely related to the previous value. The two basic measures of persistence are the autocovariance and the autocorrelation coefficient.