[R-sig-eco] Displaying estimators from specpool or estimateR

Martin Unterseher martin.unterseher at uni-greifswald.de
Tue Jan 18 09:22:22 CET 2011

Dear all,

as far as I know, functions specpool and estimateR (both in package  
vegan) provide only the final values of the non-parametric estimators.  
In contrast to that, EstimateS calculates a real curve of each  
estimator function. To me this is of importance to evaluate the  
stability and reliability of the final value calculated by the  
respective estimator.

Carsten Dormann found a way to achieve this in R and provided this  
function for me:

 > reps=100
 > sequence.abies.myk<-1:nrow(abies.myk.com) #abies.myk.com is my  
species-samples matrix from mycorrhizal fungi of Abies alba
 > res.abies.myk<-matrix(ncol=9, nrow=length(sequence.abies.myk))
 > colnames(res.abies.myk)<-names(est.abies.myk) #est.abies.myk is the  
result from specpool(abies.myk.com)
 > for (i in sequence.abies.myk)
    sample.raw.abies.myk <- replicate(reps,  
specpool(abies.myk.com[sample(x=1:nrow(abies.myk.com), size=i),]),  
    subsample.abies.myk <- matrix(unlist(sample.raw.abies.myk),  
nrow=9, ncol=reps)
    rownames(subsample.abies.myk) <- names(sample.raw.abies.myk)
    res.abies.myk[i, ] <- rowMeans(subsample.abies.myk)

It works fine. For small data sets and many replications, curves  
shapes are in concordance with the EstimateS output. For larger data  
however, calculations with high values of "reps" would take hours to  
days. If you use small values for "reps" instead the curves will  
appear jagged and are often not publishable.

Does any one know of alternatives to display estimator curves in R?


Dr. Martin Unterseher
Universität Greifswald
Institut für Botanik und Landschaftsökologie
Lehrstuhl für Allgemeine und Spezielle Botanik
Grimmer Str. 88
17487 Greifswald

Tel. 03834 / 864184
Fax. 03834 / 864114


More information about the R-sig-ecology mailing list