0
$\begingroup$

When I plot PACF graph of my data it is giving decimal numbers on X-axis. What does that mean?

0.08 lag dos it imply it depends on 0.08th segment of the day? enter image description here

$\endgroup$
5
$\begingroup$

The scale of the horizontal-axis is in terms of complete seasonal periods. It looks like you might have daily data with a seasonal period of 365. So the horizontal axis is in terms of years. 0.08 lags means 0.08 years, or equivalently 0.08*365=29 days.

$\endgroup$
-1
$\begingroup$

Let us suppose you are finding out the pacf for the time-series, series_x. Just write: pacf(series_x[1:length(series_x)]). That will give the lags in integers.

If you need to find the cross correlation between the two series, where time-series_x is series_x and time-series_y is series_y, (note that the length of series_x must be equal to the length of series_y) just write:

library(forecast) 

fit  <- auto.arima(series_x) 
yhat <- residuals(Arima(series_y, model=fit))
res  <- ccf(fit$residuals[1:nrow(series_x)], yhat[1:nrow(series_y)], 
            lag.max=12, type=c("correlation","covariance"), plot=TRUE)
$\endgroup$
  • $\begingroup$ This does not answer the question. $\endgroup$ – Stephan Kolassa Nov 27 '17 at 8:38
  • $\begingroup$ Thanks Stephan for your suggestions , the first answer here has well explained the solution of the problem . I have just extended to that if someone wants integer lags in pacf and ccf plots . $\endgroup$ – Ankit Kamboj Nov 28 '17 at 6:03

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.