After last week’s post it was commented that strucchange and segmented would be more suitable for my purpose. I had a look at both. Strucchange can find a jump in a time series, which was what I was looking for. In contrast segmented is more suitable for occasions where rates of effect. This made me think of a post on climate change I made earlier, so I applied segmented to those data

Strucchange and Nile data

Strucchange is a package to detect jumps in data. They have an example of the effect of the Aswan dam on Nile water. Those data are suitable to compare the methods available. For strucchange this is just following the example. The result is an effect on water flow around 1898. There is just one catch. It needs a time series, so regular data over time.
library(strucchange)
library(segmented)
library(R2jags)
library(tree)

data(“Nile”)
plot(Nile)
bp.nile <- breakpoints(Nile ~ 1)
ci.nile <- confint(bp.nile, breaks = 1)
lines(ci.nile)
Interestingly tree gives 1898.5, about the same answer. Segmented does not work. It is not devised for this kind of step changes.
NileDF <- data.frame(Nile=as.numeric(Nile),year=1871:1970)
tree(Nile ~ year,data=NileDF)

node), split, n, deviance, yval
      * denotes terminal node

 1) root 100 2835000  919.4  
   2) year < 1898.5 28  492000 1098.0  
     4) year < 1889.5 19  388200 1067.0  
       8) year < 1880.5 10  205200 1133.0 *
       9) year > 1880.5 9   92680  994.6 *
     5) year > 1889.5 9   48760 1162.0 *
   3) year > 1898.5 72 1105000  850.0  
     6) year < 1953.5 55  802300  836.1 *
     7) year > 1953.5 17  258600  894.7  
      14) year < 1965.5 12  114300  947.8 *
      15) year > 1965.5 5   29480  767.4 *

segmented(lm(Nile ~1,data=NileDF),~ year,psi=list(year=c(1890)))

Error in segmented.lm(lm(Nile ~ 1, data = NileDF), ~year,