A new version of palaeoSig (1-1.2) has been uploaded onto CRAN (many thanks to the CRAN volunteers who processed it so quickly). This R package includes functions to test the statistical significance of palaeoecological reconstructions made using calibration (transfer) functions, and methods to check the impact of spatial autocorrelation on calibration model performance.
This version includes a fix for a potentially serious bug in obs.cor ( ), some minor graphical improvements, and new functions that can make significance testing much faster when testing reconstructions from multiple sites.
obs.cor( ) will now work correctly if the species names in the calibration set and the fossil data are not in the same order (Thanks to Mathias for noticing this).
Both plot.palaeoSig( ) and plot.obscor( ) now allow the user to set the p-value highlighted on the graphs. The default is 0.95.
In the previous version of palaeoSig, if you wanted to test reconstructions from multiple sites, you had to run randomTF( ) for each site. Each time it ran, randomTF( ) would create many, by default 99 — ideally many more — calibration models trained on random data. With some models this is very slow, and it was inefficient to have to calculate this set of models several times.
In the new version, it is possible to split the model making from the reconstruction testing. Function ModelMaker( ) will fit models to the observed data and many random variables. The output of ModelMaker( ) can be used to test reconstructions at multiple sites using randomTFmm( ) once for each set of fossil data. With many sites, this is potentially many times faster.
library(palaeoSig) data(SWAP) data(RLGH) #old style rlgh.wa<-randomTF(spp=sqrt(SWAP$spec), env=data.frame(pH=SWAP$pH), fos=sqrt(RLGH$spec), fun=WA, col=1, n=99) #new style swap.mods<-ModelMaker(spp=sqrt(SWAP$spec), env=data.frame(pH=SWAP$pH), fun=WA, n=99)#new rlgh.wa2<-randomTFmm(fos=sqrt(RLGH$spec), modelList=swap.mods, fun=WA, col=1)
These functions do not work with the Modern Analogue Technique, as MAT lacks a separate calibration step. As always, it is the user’s responsibility to check that the modern and fossil data are transformed in the same way.