I have a raster stack/brick of 35 annual rainfall raster layers for Per-pixel temporal trends. I have calculated Sen Slope by applying Theil-Sen test, a non-parametric linear regression model using R (code is written below).Now I will have to check the significance of Theil-Sen slope using Mann–Kendall (MK) trend test at different confidence level 90, 95 and 99% accounting for serial correlation present in time series raster images.
Can someone guide me how to do this using raster brick in R?
library(sp)
library(rgdal)
library(raster)
library(Kendall)
library(boot)
library(trend)
library(zyp)
library(mblm)
getwd()
setwd("H:/arrangedlineartrends/2004") #set working directory
bundle
Didn't look deep into your code but you can view the two-sided p-value of the MannKendall Function by feeding it to the summary() function or by calling the list-value of the Kendall-object: MannKendall(...)$sl
I don't know what a raster stack is, so if that's important to your question, my comments may be less helpful.
Your question is actually unclear to me. I don't know what you mean by determining the significance at different confidence levels.
The function trend::sens.slope reports the confidence interval for Sen's slope. Unfortunately it appears the the trend::seas.sens.slope does not. See the following for an example.
http://rcompanion.org/handbook/F_13.html
In R, it is easy to get the confidence interval for any statistic by using bootstrap methods. See "Optional analyses: Confidence intervals for p-values and R-squared values" in the following.
http://rcompanion.org/handbook/G_10.html
I think the problem with trying this method with Sen's slope, is that Sen's slope is so robust to outliers that you might get a confidence interval of zero width.
Not sure if that helps.
Thanks Alexander Steiger and Salvatore S. Mangiafico for your suggestions.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Software
- R Statistical Package