Contact: Brian Cade
Staff at the Fort Collins Science Center (FORT), U.S. Geological Survey, have developed Blossom, an interactive program for making statistical comparisons with distance-function based permutation tests and for testing parameters estimated in linear models with permutation procedures. The distance tests were developed by P. W. Mielke, Jr., and colleagues at Colorado State University. The tests for parameters estimated in linear models with permutation procedures were developed by B. S. Cade and J. D. Richards at FORT, U.S. Geological Survey.
Blossom includes routines for multiresponse permutation procedures for one-way (MRPP), randomized block (MRBP), paired (PTMP),and sequential (MRSP) tests. Two options, Euclidean distance and Hotelling's variance/covariance commensuration are provided for standardizing multivariate data in the MRPP family of tests. Exact probabilities can be computed optionally for small sample designs in MRPP; Monte Carlo resampling and a 3 moments approximation are available for larger samples. Multivariate medians and distance quantiles are estimated for use as descriptive statistics with MRPP analyses. Permutation tests for 3 forms of linear models are provided, ordinary least squares (OLS), least absolute deviation (LAD), and quantile regression (LAD/QUANT). In addition to a drop in dispersion form of linear model test, permutation and asymptotic chi-square distributional approximation of P-values are provided for the quantile regression rank score test. Simultaneous estimation of all regression quantiles is an option. Empirical coverage tests for univariate goodness-of-fit and g-sample comparisons that are extensions of the Kolmogorov-Smirnov family of statistics for comparing cumulative distribution functions, including an option for testing goodness-of-fit for a random uniform distribution on a circle are included.
The computer code was compiled with Lahey/Fujitsu Fortran v7.1 (Fortran 95) Win32 compiler to dynamically allocate memory, allowing for virtual-sized data sets. Access to Windows is provided using the Winteracter GUI Toolkit (ISS, Ltd).
Note from Authors: In 2007 we added new installation and runtime capability to recognize read/write/delete privileges invoked on the computer running Blossom. This fixes some problems when first running Blossom upon installation. In addition, we included a new method to store the distance matrix used in several statistical procedures to reduce the memory storage requirements and allow larger problems to be analyzed
blossom2.exe [8.8 MB] This version is optimized to run on Windows 2000, Windows XP, or Windows Vista computers using an Intel Pentium 4 processor with SSE2 instructions. If this version does not run on your computer, please contact the authors for a version appropriate to your needs. [Updated 5.2.2008]