Hello, This is only a technical comment: I won't have no opinion on the sensibility of the analysis nor on the choice of the method.
On Tue, 2010-03-30 at 22:00 -0700, Josh Rasmussen wrote: > I have a dataset of monthly hydrologic parameters by year, including this > year, with a matrix of 48r x 24c. > > > my.data > Octinflow Novinflow . . . OctPrecip . . . > 1963 400 357 0.25 > 1964 567 429 0.89 > . > . > . > 2010 150 210 0.12 > > > I would like to use NMDS to rank years based on similarity to 2010. > > >library(vegan) > >my.model <- metaMDS(my.data, distance='euclidean', autotransform=T, k=6) > > Is it possible to use the euclidean distance of the points to determine the > overall (dis)similarity of years... > Euclidean distance sounds sensible, but 'autotransform = TRUE' does not. I suggest you have 'autotransform = FALSE' and transform data prior to analysing them. The data clearly need transformation since your variables seem to be measured in different units. The variables (columns) must be commensurable for meaningful Euclidean distances. > > dist(my.model$points) > > And then from this extract the values of interest (i.e. the last row) and > rank these values? > Technically, you can do this, but why do you need NMDS? You could do this with original data. Another point is that your observations are noisy and just comparing all sites to one noisy site gives you noisy ranking. Cheers, Jari Oksanen _______________________________________________ R-sig-ecology mailing list R-sig-ecology@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-ecology