Rarefaction analysis via ressampling
Calculates the repeatability of a statistic of the data, such as correlation or covariance matrix, via bootstrap resampling with varying sample sizes, from 2 to the size of the original data.
Rarefaction( ind.data, ComparisonFunc, ..., num.reps = 10, correlation = FALSE, replace = FALSE, parallel = FALSE )
ind.data |
Matrix of residuals or indiviual measurments |
ComparisonFunc |
comparison function |
... |
Aditional arguments passed to ComparisonFunc |
num.reps |
number of populations sampled per sample size |
correlation |
If TRUE, correlation matrix is used, else covariance matrix. MantelCor always uses correlation matrix. |
replace |
If true, samples are taken with replacement |
parallel |
if TRUE computations are done in parallel. Some foreach backend must be registered, like doParallel or doMC. |
Samples of various sizes, with replacement, are taken from the full population, a statistic calculated and compared to the full population statistic.
A specialized ploting function displays the results in publication quality.
Bootstraping may be misleading with very small sample sizes. Use with caution if original sample sizes are small.
returns the mean value of comparisons from samples to original statistic, for all sample sizes.
Diogo Melo, Guilherme Garcia
## Not run: ind.data <- iris[1:50,1:4] results.RS <- Rarefaction(ind.data, PCAsimilarity, num.reps = 5) results.Mantel <- Rarefaction(ind.data, MatrixCor, correlation = TRUE, num.reps = 5) results.KrzCov <- Rarefaction(ind.data, KrzCor, num.reps = 5) results.PCA <- Rarefaction(ind.data, PCAsimilarity, num.reps = 5) #Multiple threads can be used with some foreach backend library, like doMC or doParallel #library(doParallel) ##Windows: #cl <- makeCluster(2) #registerDoParallel(cl) ##Mac and Linux: #registerDoParallel(cores = 2) #results.KrzCov <- Rarefaction(ind.data, KrzCor, num.reps = 5, parallel = TRUE) #Easy access library(reshape2) melt(results.RS) ## End(Not run)
Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.