Suppose I have the following 20 points:

c(-0.560475646552213, -0.23017748948328, 1.55870831414912, 
0.129287735160946, 1.71506498688328, 0.460916205989202, -1.26506123460653, 
-0.686852851893526, -0.445661970099958, 1.22408179743946, 
0.400771450594052, 0.11068271594512, -0.555841134754075, 1.78691313680308, 
0.497850478229239, -1.96661715662964, 0.701355901563686, -0.472791407727934

These 20 data points are anomalies of winds or the annual deviation from the 1981-2010 mean. I extracted this from one grid point in a gridded data set (i.e., in netcdf format). For simplicity, I am only showing one time series.


I am analyzing extreme heavy rainfall days and to understand the formation mechanism, I have to perform lag compositing.

First, I derived the anomalies of the 20 years relative to climatology (daily value minus the climatological value). For example, the 20 years is from 1981-2000 while the climatology is the average from 1981-2010.

In each year from 1981-2000, I derived the days exceeding a certain threshold. This is labeled as Lag 0. The day before this is Lag-1 and the day after is Lag +1. This is a common method of compositing.

Each lag is the average of 20 points.

What I want

I would like to test for the LOCAL significance of these anomalies. For example, I want to test the significance of the probability of occurrence of these extreme cases. Like a Monte Carlo simulation.

I would like to ask for suggestions for a correct statistical test for this and how should I do the test. I want to use R in performing this test.

I'll appreciate any help.

Sincerely, Lyndz


I would suggest looking at the detection of Additive Outliers in Time Series.

Here is a possible reference: https://msed.vse.cz/msed_2014/article/261-Marek-Lubos-paper.pdf


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.