[R] Calculating Kendall's tau

[David Winsemius]

On Apr 2, 2015, at 7:06 AM, S Ellison wrote:

>> I am analyzing trend  using Mann-kendall  test for 31 independent sample, each
>> sample  have 34 years dataset.  I supposed to find Kendall “tau” for each
>> sample. The data is arranged in column wise (I attached  the data).To find
>> Kendall tau, I wrote R script as:
>> ...
>> Anyone can tell me how can I get orderly displayed  “tau” value? 
> 
> Usually, in R, a hypothesis test returns an object, and you can extract an individual element of that object.
> 
> MannKendall seems to be no exception. Looking at the help page, a MannKendall test returns...
> " A list with class Kendall.
> tau 	Kendall’s tau statistic
> sl 	two-sided p-value
> S	Kendall Score
> D	denominator, tau=S/D
> varS	variance of S"
> 
> To get just tau, say something like MannKendalltau[i]<-MannKendall(y[,i])$tau
> 
> But your code is a bit of a mess....
> MannKendalltau<- numeric(nc) simply makes MannKendalltau a single integer equal to nc; that doesn't look sensible when the next thing you do is treat MannKendalltau as a vector. R's been kind to you and extended MannKendalltau when you tried to add things to later, non-existent, elements, but it clearly wasn't the right thing to do. Look up ?numeric, and then look up ?vector for next time you want to set up an empty vector.
> 
> Second, since MannKendall(y[,i]) ) returns a list object of class Kendall, MannKendalltau[i]<-MannKendall(y[,i]) assigns a whole  object containing 5 values to each new element of your MannKendalltau. So your result is a list of lists.
> 
> Finally, you don’t need a loop at all. On a data frame, sapply would work nicely, so (although I've not tested it) something like
> 
> sapply(desta[,2:nc], 2, function(x) ManKendall(x)$tau)

That looks more line an apply call. The second argument to sapply needs to be a function or function name.


-- 
David.
> 
> ought to do the whole thing in one shot and package it nicely into a named vector while it's about it.
> 
> S Ellison
> 
> 
> *******************************************************************
> This email and any attachments are confidential. Any u...{{dropped:19}}