I have several small biological count-based data sets with one or more rows having zero proportion. The other proportions in the row sum to 1.000 (or 0.9999 in the sixth data row below because of rounding errors in the computer). An example is:
sampdate filter gather graze predate shred 2000-07-18 0.0550 0.5596 0.0734 0.2294 0.0826 2003-07-08 0.0734 0.6147 0.0183 0.2294 0.0642 2005-07-13 0.1161 0.5714 0.0357 0.1696 0.1071 2006-06-28 0.1000 0.4667 0.1500 0.1333 0.1500 2010-09-14 0.0778 0.6111 0.0444 0.1889 0.0778 2011-07-13 0.0879 0.5714 0.0659 0.2747 0.0000 2012-07-11 0.1042 0.5313 0.0625 0.2396 0.0625 My concern is that in most field-biological (ecological/environmental) data there can be two explanations for zero counts: the organism was not present on that date or it was present but not collected. There is no way to determine which case holds true in each instance, but the ecological interpretations differ. The zCompositions package offers several methods of imputing a value to replace the zeros. As I'm completely new to compositional data analyses (CoDA) I would appreciate advice on how to select the most appropriate method for these data sets. The available methods are: Geometric Bayesian multiplicative, BM, (GBM, default); square root BM (SQ); Bayes-Laplace BM (BL); count zero multiplicative (CZM); user-specified hyper-parameters (user). These biological data seem to me to be different from geochemical or economic data I see in package data sets or the CoDA references I've acquired and read. Advice and suggestions (including references to application of CoDA to ecological/environmental data) will be appreciated. Rich ______________________________________________ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.