0
$\begingroup$

I have a time series data-

53.97 63.32 57.06 60.27 69.46 75.08 78.31 73.28 85.84 69.34 62.57 60.11 55.63 47.29 61.22 58.46 66.26 59.71 51.12 39.36 51.89 53.85 47.64 31.75 35.01 31.34 26.64 29.48.........

Are there any set steps to approach any time series or should I just look at statistical property of data that choose from available econometrics model like ARCH, AR, MA, ARIMA ?

$\endgroup$
  • $\begingroup$ Interesting side-issue of what is econometric. ARIMA for example may be useful in econometrics, but it was popularised by mainstream statisticians. $\endgroup$ – Nick Cox Apr 15 '16 at 16:00
1
$\begingroup$

Arpit,

Sometimes you need to consider Determinstic variables before you consider ARIMA variables.

I ran this model using two time trends (ie 0,0,0,1,2,3,4,etc) using Autobox(I am a developer) an AR1 and found one outlier at period 9.

enter image description here

enter image description here

$\endgroup$
  • 2
    $\begingroup$ So, basically, you're recommending OP to use autobox, right? Otherwise, I don't see how is this an answer to OP's question $\endgroup$ – Aksakal Apr 15 '16 at 15:40
  • $\begingroup$ The outlier at 9 is an outlier presumably in the sense that it is awkward to fit to this kind of model, e.g. associated with large absolute residual. Otherwise it looks like an inlier to me! $\endgroup$ – Nick Cox Apr 15 '16 at 16:02
  • 2
    $\begingroup$ "This is not written in any text book". Hard to verify! But what precisely does "deterministic variable" mean here? If it means that a model is a mix of deterministic and stochastic elements, then that idea is standard in many places! $\endgroup$ – Nick Cox Apr 15 '16 at 16:14
  • 2
    $\begingroup$ Thanks; I see now. The outlier corresponds to a small residual because it was accounted for ad hoc by a pulse term. $\endgroup$ – Nick Cox Apr 15 '16 at 16:20
  • 2
    $\begingroup$ Please do remember to disclose your affiliation with this product as per stats.stackexchange.com/help/promotion. $\endgroup$ – Scortchi Apr 15 '16 at 19:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.