Dear list, first of all, I'm not a professional researcher, but a german hobbist ;-)
Ok, I have a data set (n=75) of some morphological characters (x=15) containing both, quantitative (e.g. scale counts) as well as qualitative characters (binary coded 0=absent; 1=present). What I need to do is to a) check the significane of each character (can I really use it in morphometric analysis) and b) verify my assumption that there is more than one (undescribed) taxon involved (I have created OTUs based on distribution and on current knowledge of the species, but I would need to provide evidence for how I constructed the OTUs). I though of using non-metric MDS to verify my constructed OTUs, due to the fact, that there are no initial assumptions to be tested, and that NMDS is also able to "handle" binary coded characters. I would first natural log-transform the raw data and subtract the mean (or would it be better to use standard normalizsation [mean=0; derivation=1]). I would then run a CVA on the log transformed data set to visualized the morpholgical distinctions of each OTU. ANOVA requiers normal distributed data, but binary coded data are a priori never normal distributed, so what can I use to test significance of the characters? But I'm not sure if that's the right way to do. Any suggestions? I don't want to get killed by the reviewers ;-) btw.: I'm using the PAST 1.34 program... Thanks in advance, Wulf -- Replies will be sent to the list. For more information visit http://www.morphometrics.org
