Finding patterns in data

Ancient DOS PATN vs PATN v3

The original DOS version of PATN contained a huge suite of options, but most were rarely used. The Windows version has fewer options but is far simpler to use. The underlying algorithms in PATN V3 are basically the same as the DOS version of PATN, but the user-interface of PATN V3 provides a radically different analysis environment. For those unusual people who occasionally ask, the following table outlines the differences in utilities and options between the DOS version of PATN and PATN V3.

DOS PATN Description PATN v3
ALOB/C Non-hierarchical clustering/classification Yes*
ASIM ANOSIM (randomization tests based on association) Yes
ASO Association measures between rows (17) Yes 4
ASON Input, manipulation and output of association matrices Yes (LSM)
BOND Bonding lists based on NNB No
CHI2 Chi-squared statistic across groups No
COLR Mapping of objects by colour Yes
DATN Data input, editing and export Yes
DEND Dendrograms (from hierarchical classification) Yes
DCOR Detrended correspondence analysis (DECORANA) No
FUSE Agglomerative hierarchical clustering (8) Yes 4
GASO Association between rows with variable grouping No
GDEF Define or manipulate group definitions (9) Yes 3
GSTA Statistical summaries of variables across groups Yes
HIST Histograms and univariate statistics Yes
LABN Input, creation and output of labels Yes
MASK Masking, sampling of data Yes
MCAO Monte-Carlo testing of variables in ordination Yes
MDIV Monothetic divisive clustering No
MERG Left-right and up-down merging of data Yes
MST Minimum spanning tree Yes
NNB Nearest neighbour analysis No
PCA Principal component/coordinate analysis No
PCC Fit variables to ordination using multiple linear regression Yes
PCR Orthogonal rotation of ordination axes No
PDIV Polythetic divisive clustering No
PRAM Specify data and environmental parameters Yes
PROC Procrustes analysis (compare two ordinations) No
RAND Data generation by random variates (8 distributions) Yes
RIND Hubert & Arabie statistic of group difference No
SALE Travelling salesman network No
SAMP Row or column sampling strategies No
SCAN Presence/absence statistics Yes
SCAT Scatter plots Yes (ord)
SENS Sensitivity/ redundancy analysis of variables across groups No
SERE Seriation (1d ordination by 'smoothness') No
SSH Semi-strong hybrid multidimensional scaling Yes
TRNA Transformations and standardizations of association matrices No
TRND Transformations and standardizations of data (12) Yes 7
TSPN Pre-processor for Acrospin 3d vizualization (+SPIN) Yes better!
TWAY two-way table (re-ordering of data by classifications) Yes
TWIN Mark Hill's Twinspan (two-way indicator species analysis) No

*PATN v3+ does not include ALOB's weighting of variables option as such, but a weighting may be achieved using a combination of variable standardization and Minkowski metrics. PATN V3 does handle large datasets (1 million+ objects).