1
$\begingroup$

I am trying to understand what was done with timeseries from one paper.It says

The raw time series are prewhitened, typically by convolution with an autoregressive
filter, to reduce the spectral dynamic range. The autoregressive filter must be
computed robustly;

I have my ts x,how to perform convolution with autoregressive filter in R?Could library forecast be used or...?

EDIT What I have done so far

d <- scan('262_V01_C00_R000_TEx_BL_128H.dat')
dts <- ts(d,frequency=32)

png("n1.png")
plot.ts(dts)
dev.off()

Then I have plot enter image description here

If I perform two-time differencing

d <- scan('262_V01_C00_R000_TEx_BL_128H.dat')
dts <- ts(d,frequency=32)
dts1 <- diff(dts, differences=2)

png("n2.png")
plot.ts(dts1)

Then my plot looks like this enter image description here

Should I preprocess the data again or not before I start building ARIMA models?

Improve this question
$\endgroup$
3
  • $\begingroup$ I would probably adjust/modify/change the data points that are creating the visually unusual values in your first graph. $\endgroup$
    – IrishStat
    Commented Dec 19, 2016 at 20:40
  • $\begingroup$ @IrishStat But how to do this? $\endgroup$
    – MikiBelavista
    Commented Dec 19, 2016 at 21:48
  • $\begingroup$ I would replace each exceptional value with a local average (excluding the unusual value) $\endgroup$
    – IrishStat
    Commented Dec 19, 2016 at 21:56

1 Answer 1

Reset to default
2
$\begingroup$

Identify an ARIMA filter while taking account (robust) any deterministic time trends, level shifts, seasonal pulses or 1 time anomalies. The idea here is transform the stationary X into a white-noise series. This same filter will then be used on the stationary Y series. The subsequent cross-correlation analysis between these two filtered series yields clues to the form of the transfer between the original X and the original Y. See here for more.

Improve this answer
$\endgroup$
1
  • $\begingroup$ Please,take a look of my edit. $\endgroup$
    – MikiBelavista
    Commented Dec 19, 2016 at 19:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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