0

I am performing an interrupted time series analysis. I plotted the data and then checked for autocorrelation using ACF and PACF.

Here is the ACF on LHS and PACF on RHS

par(mfrow=c(1,2)) 
acf(residuals(model_ols), main="ACF") 
acf(residuals(model_ols), type = "partial", main="PACF")

How do make sense of this plot now? Can someone tell me what should be the next step?

Polime
  • 1
  • 1
  • 1
  • i am using the following code: par(mfrow=c(1,2)) acf(residuals(model_ols), main="ACF") acf(residuals(model_ols), type = "partial", main="PACF")...There are 16 observations . I hope lag.max is fine. – Polime Jul 12 '18 at 18:24
  • 1
    I would judge there's basically nothing going on here. The fact that PACF(2) (and ACF(4) is marginally beyond the 95% confidence interval is compensated by the fact that you're looking at about 25 comparisons ... – Ben Bolker Jul 12 '18 at 18:54
  • @BenBolker when I use p=2 and then check for model improvement for p=3 or 4 through anova(model_p2,model_p3), it shows very significant p-value. What can I do now?? – Polime Jul 12 '18 at 19:14

0 Answers0