Hello all, I have a general question about time series and I wonder if someone could help me. I have time series data of this form: x=c(rnorm(500,0,1),rnorm(500,5,1),rnorm(500,10,1),rnorm(500,3,1),rnorm(500,8,1),rnorm(500,4,1),rnorm(500,1,1),rnorm(500,7,1)) time=1:4000 plot(time,x) Each "rnorm" generates a different cluster. I would like to do a statistical test of mean difference between the first, the second and the seventh cluster: HO: mean(rnorm(500,0,1)) = mean(rnorm(500,5,1)) = mean(rnorm(500,1,1)) My questions are: (a) Can I simply do this by an ANOVA model on these values or by change-point analysis (using "multiple.mean.norm" of the package "changepoint")? (b) If I want to check for the autocorrelation of my series (by "acf"), which data should I use: the residuals of the ANOVA model or the actual data? The actual data will give me high autocorrelation due to the trend, is that correct? Thank you for your help, Mike