PAST version history

Version 1.00:

• Improved Shapiro-Wilk test (max N increased from 50 to 5000)
• Improved Rayleigh test for directional data (max N increased from 50 to 200)
• Hamming distance for categorical data added to cluster analysis, PCO and NMDS
• Permutation test added to "compare diversities" module
• Permutation test added to chi-squared module

Version 1.01: Plotting on axes 2+3 in DCA. Eigenvalues 1-4 reported.

Version 1.02: RMA now works properly also for negative slopes.

Version 1.03:

• 95% confidence interval for autocorrelation
• Black square when copying graphic, bug fixed
• Seriation copy data, bug fixed
• Grid lost when resizing multivariate allometry graphic, bug fixed
• XY graph: added support for row colors and labels

Version 1.04: RMA: Standard error for the intercept used an incorrect formula in Davis (1986). Ver. 1.04 follows Miller & Kahn (1962).

Version 1.05

• Click on any graph for setting grid, axes and axis labels
• Convex hull option added to all scatter plots
• Correlations between variables and PCs added to PCA loadings plot
• Data plotted as points or curve in Sinusoidal Fit
• Max number of sinusoids increased from 6 to 8 in Sinusoidal Fit
• Algorithm for Sinusoidal Fit more robust (max iterations in SVD increased)

Version 1.06

• Bootstrapping of linear regression/RMA
• One-sample t test versus theoretical mean
• Continuity correction in Mann-Whitney test
• Bug fixed: Crash when quitting PAST for some Windows configurations

Version 1.07

• ANOSIM (Analysis of Similarities) with 12 distance measures
• Stress value in Shepard plot (Non-metric MDS).
• Taxonomic distinctness: Problem with data not starting in first row fixed.

Version 1.08

• Cross-correlation with p values
• PCA can now optionally use the SVD algorithm

Version 1.09

• Mahalanobis-distance permutation test for two multivariate groups
• Removed limit of 210 items in cluster analysis

Version 1.10

• Black-and-white option for all graphics
• Y axis can be scaled in cluster analysis

Version 1.11

• Normal probability (normal QQ) plot with RMA regression
• 50% speed increase in Ward's method clustering
• CVA loadings (eigenvectors)
• Affine component (0th partial warp) scores for thin-plate spline deformations.

Version 1.12

• NPMANOVA (Non-Parametric MANOVA) with 12 distance measures
• 'Edit labels' mode in the data entry spreadsheet
• Copy data from the F/t and chi^2 test windows
• Bug fixed: Occasional crash in PCO scatter plot
• Bug fixed: CVA loadings window not staying in front

Version 1.13

• Simple implementation of the CONOP (Constrained Optimization) method for biostratigraphy
• Reporting of residuals from linear regression

Version 1.14: Added numerical output ("View numbers") to PCO, CA and DCA.

Version 1.15

• User selection of pairs of landmarks to link with lines in landmark, thin-plate spline, partial and relative warp and shape PCA plots.
• Optional row labels in 3D landmark plot
• Warning when quitting PAST without saving changes
• Automatic scaling when reading TPS files (SCALE keyword in file)
• "Dots for blank fields" bug when copying data from PAST to other programs fixed.

Version 1.16

• Partial warp deformations went opposite way of scores. Fixed.
• Procrustes fitting (2D) occasionally mirrored some configurations. Fixed.
• When copying graphics with zoomed axes, elements from outside the window could occasionally be included. Fixed.
• When copying graphics from small windows, curved lines could become ragged. Partially fixed (by rounding instead of truncation).

Version 1.17

• CONOP: Occasional long pause at startup, bug fixed
• CONOP: Option for using event numbers instead of names in range chart
• RASC: PAST can now read files in the RASC .DAT-format
• RASC: Reports the number of cycles
• RASC: Dendrogram truncation at bottom and top, bug fixed
• RASC: Loss of taxa when rerunning RASC without closing window, bug fixed.

Version 1.18

• "Commatizing" made unnecessary and removed
• View PCA loadings as numbers
• Centering option in PCA scatter plot
• Faster eigenvalue routine

Version 1.19

• Paired Hotelling's T² test
• Two-way cluster analysis (combined Q and R mode)
• Support for missing data added to more functions
• Raup-Crick and Horn indices added to MDS and PCO
• Reporting of all eigenvalues in PCA (not only first four)
• Unitary Associations: Added support for marker beds, substantial speed increase in cycle resolution, status label.
• Histogram: Added ML fit to normal distribution
• Univariate statistics: Added standard error of the estimate of the mean
• Edit labels: Clicking in label cell now makes cell active
• Cancel button in "Evaluate expression" fixed
• Added "Normal" (Gaussian random numbers) to Evaluate expression

Version 1.20

• Fixed two bugs in cladistics: Wrong tree length sometimes given when changing tree in cladogram display, and wrong character states sometimes given for terminal taxa when the root state is a question mark.
• Hotelling's T² made more robust using Moore-Penrose generalized matrix inverse.
• Linear regression residuals includes rows with question marks
• Experimental feature: Stratigraphically constrained cluster analysis

Version 1.21

• Raw similarity/distance matrix for all measures
• Cosine similarity measure
• Maximum number of columns for PCA (and rows for PCO) increased from 200 to 512
• Maximum number of characters for cladistics increased from 128 to 200
• Maximum number of taxa for cladistics increased from 255 to 511
• Improved reading of trees from NEXUS files
• Crash when transposing matrices with more columns than rows, bug fixed

Version 1.22

• Minimal spanning trees for PCO and NMDS
• Data set size limit removed for PCA and PCO (but less than 500 columns for PCA and 500 rows for PCO still recommended because of computation time).
• Faster Open of very large data files
• Faster Transpose of very large data sets
• "Insert more columns" occasional freeze: Bug fixed
• Diversity indices - bad behaviour for all-zero columns: Bug fixed
• Occasional missing lines in multipage dendrograms: Bug fixed
• Occasional incorrect results in polar Fourier shape analysis (not Elliptic Fourier): Bug fixed

Version 1.23 (bug fixes)

• Convex hulls: Crash when points had identical positions - bug fixed
• PCA: Eigenvalues not always in descending order when using SVD - bug fixed
• Graph: Drawing area blocked out with color for some graph setting combinations - bug fixed
• Diversity curve: Crash when viewing numerical output - bug fixed
• Cladistics: Improper handling of characters with identical character state for all taxa - bug fixed
• DCA: Crash if a column had all zeros - bug fixed
• Ternary plot: Added option for row labels

Version 1.24

• Procrustes fitting (2D): Option for rotation to major axis
• New function for rotating and flipping landmark configurations (2D)
• Directions/rose plot: New "orientations" (0-180 degrees) option
• Cladistics: bug fixes for large data sets (now limits of 200 taxa and 400 characters)

Version 1.25

• Added offset value to discriminant function output
• PCO: Added plotting of axes 2+3
• NMDS: Factor 10 speed-up
• Unitary Associations: Factor 10 speed-up for some data sets

Version 1.26

Important change in PCA: When using PCA with the correlation matrix, the PCA scores (scatterplot) was previously computed with respect to original (raw) variables. From version 1.26, the scores are computed with respect to standardized variables.

Version 1.27

• Buzas & Gibson's evenness e^H/S added to 'diversity indices' and 'compare diversity indices'.
• Cluster analysis: Option to save dendrogram in NEXUS format (can be viewed with e.g. TreeView)
• Font name and size added to graphic preferences
• Improved algorithm for automatic selection of plotting axis ranges
• Very rare crash in the 'Univariate statistics' module, bug fixed.

Version 1.28

• Fixed bug introduced in ver. 1.27: Crash in some graphic windows
• PCA scattergram/biplot: Plotting on axes 1+2, 1+3 and 2+3

Version 1.29

• Added Manhattan distance to distance measures, cluster analysis, PCO, MDS, ANOSIM and NPMANOVA
• Fixed minor problem with placement of some windows under RASC
• Added variable "stdev" (standard deviation) to "Evaluate expression"

Version 1.30

• "Capture-recapture" species richness estimators: Chao2, first- and second-order jackknife, bootstrap.
• Size removal of multivariate morphometric data by Burnaby's method
• Bootstrapped confidence intervals in logistic fit
• Option to standardize rotation and starting point in Elliptic Fourier Analysis
• Increase from 9 to 12 different symbols/colors

Version 1.31

• 95% Working-Hotelling confidence bands for linear regression line
• All negative values used to be interpreted as missing data for computation of similarity indices (in cluster analysis, PCO, MDS, ANOVA, NPMANOVA etc.). Now only the value -1 (or '?') is accepted as coding for missing data.
• Improved printing, especially for non-postscript printers
• Bug fix: New symbols/colors in ver. 1.30 were not included in clustering labels, confidence ellipses and convex hulls.
• Logistic bootstrap: Better error reporting

Version 1.32

• Gridding (spatial interpolation): Kriging with semivariogram modelling, thin-plate spline, moving average.
• Matrix plot (grayscale)
• Spindle diagram (fossil abundance through stratigraphy)

Version 1.33

• Gaussian mixture analysis with EM algorithm
• Bug fix in Correspondence Analysis (Detrended Correspondence Analysis not affected): Occasional flip of Q mode axes relative to R mode
• Automatic linear trend surface removal in gridding module
• "Massage" menu changed name to "Transform"

Version 1.34

• User-specified number of dependent variables (constraints) for Hotelling's T² and MANOVA
• Contouring in gridding and matrix plotting (slightly buggy on boundaries)
• Option to set symbol size in all plots
• "Sample variance" and "Sample standard deviation" in Univariate statistics caused confusion - removed.
• Bug fixes in loading of Nexus tree files

Version 1.35

• Two-way ANOVA
• Mardia's multivariate normality tests (multivariate skewness & kurtosis)
• Option to change tick mark density in all plots
• Cross-evaluation (leave-one-out) in discriminant analysis
• Choice of allometric or isometric size removal in Burnaby module
• Box's M: Better error reporting

Version 1.36

• Canonical correspondence analysis (CCA)
• Bookstein landmark registration
• Distance/similarity measures: Full symmetric matrix given, not only upper triangle
• Minor bug fix in kriging: Crash for some datasets

Version 1.37

• Minimal spanning tree added to XY plot
• User-supplied symmetric distance or similarity matrix option in cluster analysis, PCO, NMDS, ANOSIM and NPMANOVA
• Option to set number of permutations in permutation t test, multivariate permutation test, ANOSIM and NPMANOVA
• "Sort ascending" and "Sort descending" now sort all rows based on the first column (previous versions sorted columns independently)
• PCO scores automatically scaled using square roots of eigenvalues
• Bug fix in Unitary Associations: Rare failure to detect residual virtual edges

Version 1.38

• All eigenvalues reported in PCO (not just first four)
• Bug fix in DCA: Crash for zero eigenvalues
• Added "Slope" and "intercept" captions to linear regression
• Bug fix: "Oval" and "Star" row symbols not properly loaded from datafile
• Bug fix in PCO: Numerical output crashed for zero eigenvalues
• Default XY plot mode changed from lines to points
• Point distribuion statistics: Small correction in expected distance calculation. Added option for edge effect adjustments, and option for rectangle rather than convex hull for area.

Version 1.39

• Appearance Event Ordination (biostratigraphic method)
• Univariate statistics now works for several columns
• Point distribution statistics: Output of minimal spanning tree segment directions (for testing linear arrangement)
• Directions module: Bug in chi-squared test (introduced in ver. 1.24) fixed. Rayleigh's test now handles orientations
• Detrended Correspondence Analysis: User control on number of detrending segments
• Crash when reading large TPS files, fixed

Version 1.40

• Bootstrapping in cluster analysis
• Improved user interface for the "Evaluate expression" function

Version 1.41

• Faster 3D procrustes fitting

Version 1.42

• XY plot with error bars
• Labels in linear regression plot
• Number of bins for chi-squared in directional analysis can now be set by user
• Point distribution statistics now outputs all distances and orientations between nearest neighbours (change from v. 1.39).
• Bug fixed: Raup-Crick similarity index caused occasional crash in several modules
• Bug fixed: Sorting parts of matrix did not work

Version 1.43

• Post-hoc analysis for MANOVA, by simple Bonferroni-corrected pairwise Hotelling's T²
• Mantel correlogram and Mantel periodogram for multivariate time series
• Function for regular spacing of irregular time series by interpolation
• Bug fixed: Kolmogorov-Smirnov always reported p=1 for very large sample sizes
• Bug fixed: Mann-Whitney crashed for very large sample sizes

Version 1.44

• Post-hoc analysis for Kruskal-Wallis (non-parametric ANOVA)
• Welch's ANOVA for groups with unequal variances
• Bug fixed: Incorrect Bonferroni correction for MANOVA post-hoc

Version 1.45

(Major internal revision of the code)

• Speed increase in many modules
• Large cluster analysis output no longer split into pages
• Bug fixed: Incorrect Kruskal-Wallis posthoc when data not starting in first column
• Bug fixed: Crash in MANOVA post-hoc for small sample sizes
• Bug fixed: Losing graph settings when a plot is updated
• A number of small improvements in the user interface
• Additional (experimental) corrections in diversity curve module

Version 1.46

• Fixed bugs in Transform menu functions introduced in version 1.45

Version 1.47

• Scripting language (early version!)
• DCA: Pop-up labels when moving mouse over points
• Undo levels increased to 10

Version 1.48

• Almost 30 new functions available in the scripting language
• Bug fixed: Confidence ellipses and convex hulls incorrectly drawn for datasets with same color used in several groups
• NMDS scatter plot automatically rotated to major axis

Version 1.49

• Nucleotide codes (A, T, U, C, G) accepted for all analyses
• Interleaved Nexus files accepted
• Genetic distance indices in PCO, NMDS, clustering, ANOSIM, NPMANOVA: Jukes-Cantor and Kimura
• NMDS automatically runs 10 times with random starting values, and the best solution is selected
• NMDS scatter plot axes automatically flipped according to input order
• PCO made robust to variates with zero variance

Version 1.50

• Neighbour joining clustering
• Minor bug fixes

Version 1.51

• Redo function (undo an undo operation)
• "Optimal" histogram bin size suggested automatically
• New functions and constructs in the scripting language

Version 1.52

• Mantel test
• Option for showing outliers in box plot
• Minor bug fixes

Version 1.53

• Tests for paired data: t test, sign test (exact) and Wilcoxon signed rank test (normal approximation and Monte Carlo).

Version 1.54

• New similarity measures: Kulczynski and Ochiai
• Confidence ellipses drawn with Bezier curves (graphical quality improvement)
• "Invisible rectangles" in copied graphics removed
• Files can be opened by dragging from desktop
• Cladistics: Bug fix in character reconstruction with Wagner optimisation (tree topology was OK)
• Option for showing branch lengths on parsimony trees
• Correspondence analysis: All-zero rows and columns placed on the origin in the plot (previously an error message was given)
• Some similarity measures (Dice, Jaccard etc.) were undefined for all-zeros data. This caused problems when bootstrapping sparse data sets in clustering and neighbour joining. Similarity is now set to zero in such cases.

Version 1.55

• Bug fix: Monte Carlo Wilxocon paired test gave too high p values (bug introduced in 1.53)
• Bug fix: Rare crash when plotting confidence ellipses and convex hulls
• Bug fix: Plots not scaled correctly to confidence ellipses (bug introduced in 1.54).

Version 1.56

• Option for outgroup rooting in neighbour joining cluster analysis
• Levene's test by medians in ANOVA (previously only Levene's test by means was available)
• Minor bug fix concerning colors of confidence ellipses

Version 1.56b

Fix of bug in 1.56 causing crash in several modules.

Version 1.57

• Ripley's K statistics for point patterns
• Bubble plot for 3D data
• Linear fit: Permutation test on R²
• Permutation test for Spearman/Kendall non-parametric correlation
• Correlation matrix: Spearman/Kendall option added
• Sort on color
• CA, DCA and PCO: Scatter plot on axes 1+3
• All scatter plots have mouse-over labels
• Mixture analysis: Akaike information criterion added (for selecting number of groups)
• New algorithm for setting tick marks in plots

Version 1.58

• Sinusoidal fitting can now also optimize frequencies
• Chi-squared and Akaike statistics in sinusoidal fitting
• Counter window for microfossil counting
• Linear fitting: Fix of bug in bootstrapping, introduced in v. 1.57
• Area used for Ripley's K can now be set manually
• New options for input format in Appearance Event Ordination

Version 1.59 (bug fixes)

• Linear fit: Incorrect permutation test when values were log-transformed (bug introduced in 1.57)
• Linear fit: Incorrect residuals (introduced in 1.57)
• Plotting of lines (many modules): Distortions when line clipped against lower edge (introduced in 1.54)
• Rectangles in histograms etc. not always clipped against lower edge
• Improved error messages in Mixture analysis

Version 1.60

Time series analysis: Maximum-likelihood ARMA modelling, and Box-Jenkins intervention analysis.

Version 1.61

• Pairwise post-hoc comparisons in ANOSIM and NPMANOVA
• Two-way cluster analysis plot improved, and size limit removed
• Rare crash in sinusoidal fit fixed
• New definition of Akaike information criterion in sinusoidal fit
• Correspondence analysis: All eigenvalues reported, not just four
• Rayleigh test for directional data: Limit of N=200 removed
• "p=0" changed to "p<1/number_of_permutations" in permutation tests
• Slight statistical improvement in random number generator
• Operators +, -, * and / in script language now accept mixed types such as array+double
• Default number of permutations in t test, ANOSIM and NPMANOVA increased to 10,000

Version 1.62

• Faster save
• Improved contour plots (matrix plot and gridding)
• Colour map option in matrix plot and gridding
• Option for thick lines in all plots
• Rose plot: Option for circular "grid"
• Univariate statistics: Sum of values included
• Polar and elliptic Fourier shape analysis: Number of harmonics increased to 20
• PCA: The SVD option is now default
• Script language: Decimal point/comma issue resolved
• Log (transform menu): Constant no longer automatically added for zero or negative data.

Version 1.63

• Mixed similarity/distance measures in cluster analysis, PCO, NMDS, ANOSOIM/NPMANOVA etc.
• Fisher's exact test for 2x2 contingency tables
• Jarque-Bera test for normal distribution (in addition to Shapiro-Wilk)
• Changed scale factor in PCO from sqrt(2)/2 to 1
• Support for outline and curve data types in TPS files
• A number of minor improvements and speed increases

Version 1.64

• A number of bug fixes (most of them introduced in 1.63)
• Confidence intervals in t tests
• Row colors increased from 12 to 16
• All geometric plots are forced to unit aspect ratio
• Replace function can search for empty cells (use "empty")
• Paired tests: Missing data supported by deletion

Version 1.65

• Procrustes fitting now according to Rohlf & Slice (1990)
• Solar forcing from 100 Ma to Recent (Laskar et al. 2004)
• Kernel Density Estimation in histograms
• 3D scatter/landmark plot improved
• Graphic size in pixels can be set by user
• PCA: SVD and eigenalanalysis now report eigenvalues in the same scale
• Curves made smoother using Bezier splines in a number of modules
• Width of each column can be set by dragging with mouse
• Bug fix: Imperfect printing of graphics in a number of modules
• All geometric plots are forced to unit aspect ratio

Version 1.66

• Percentile plot
• Eigenvalue and biplot scalings in PCA are now according to Legendre & Legendre
• Bug fix: Distortion of scatter plot for PCA on the correlation matrix with large datasets
• Images can be saved as file in the EMF format
• Cluster analysis results can be saved as text lists
• CONOP: Data size limits removed, and bug fix (suboptimal solutions found in some cases)
• Unitary Associations speed-up, and bug fix (residual virtual edges not properly inserted in very rare cases)
• CCA bug fix (variable labels incorrect in triplot if data not taken from first column in spreadsheet).

Version 1.67

• Sample-based species accumulation curves ("Mao tau")
• Open and save Excel files
• Shape PCA can be visualized as relative warp grids
• Two-way ANOVA now supports unbalanced designs (unequal sample sizes)
• Box plot now includes column labels
• Geometric mean in Univariate Statistics
• Open files in the CONOP format
• Filter stratigraphic events on type (FAD, LAD etc.)

Version 1.68 (bug fixes)

• Fixed incorrect interaction term in two-way ANOVA with very unequal sample sizes (introduced in ver. 1.67)
• Scale factor in TPS files now applied to all specimens even if specified only for first specimen
• Unitary Associations: Fixed crash for degenerate data sets with only one residual horizon
• Fixed reading of interleaved NEXUS files with non-Windows line breaks

Version 1.69

• Abundance models: added "broken stick" model, chi-squared values, linear-axis plot.
• Cladistics: added phylogram plot (scaled to branch lengths)
• All scatter plots: added option for centered axes through the origin
• Sample rarefaction: rare crash fixed
• Spline interpolation now supports several points with same x value

Version 1.70

• Bug fix in Elliptic Fourier Analysis: incorrect rotations for some data sets when using "Invariant to rotation" option.
• Partial warps: affine component now included also in the numerical output

Version 1.71

• SIMPER (Similarity Percentage analysis of ecological assemblage dissimilarity)
• PCA: bootstrapped confidence intervals for both eigenvalues and loadings. Confidence intervals and Broken Stick random model in the scree plot.
• Bug fix: crash in Horn, Ochiai and Kulczynski similarities for all-zero rows

Version 1.72

• Polynomial regression
• New in directional statistics: Angular kernel density estimation, confidence interval on angular mean, plot of mean with CI.
• Bug fix: Incorrect p values in NPMANOVA post-hoc tests for Euclidean, Horn and Manhattan distance (introduced in ver. 1.61)
• R² and p values in sinusoidal fitting
• Bug fix: Minimal spanning tree of only two points crash
• XY plot for several pairs of columns now support unequal N
• Chi-squared test for normal distribution removed from menu and added to Normality module
• Output from two-way ANOVA can now be copied
• Detrending option now ticked by default in DCA

Version 1.73

• 3D surface plot
• Color scale bar in matrix and gridding plots
• ANOSIM: R values for the pairwise comparisons, and support for groups with one sample
• Faster copy of large numerical data sets
• Bug fix in chi-squared test: Incorrect copying of text
• Bug fix in ternary plot: Column labels lost when copying graph

Version 1.74

• Genetic sequence statistics
• New distance measure: Tajima-Nei (for sequence data)
• NMDS: Extra trial using PCO as initial condition
• NMDS bug fix: Stress value was not always taken from the best trial
• ANOSIM+NPMANOVA bug fix: Incorrect pairwise (post-hoc) results for user-defined distance/similarity matrix with more than two groups
• CCA: Eigenvalue scaling for triplot removed, and improved convergence

Version 1.75

• One independent, several dependent variates regression with MANOVA
• NMDS now has option for three ordination axes
• Gradient species packing modelling: Gaussian fit with least squares or weighted average
• Reconstruction of environmental parameters from abundance data and species packing model, using weighted averages ("calibration")
• Directional statistics: Rao's spacing test
• Multivariate normality: Doornik & Hansen omnibus test
• SIMPER now supports more than two groups
• Improved algorithm for correspondence analysis by eigenanalysis (DCA module not affected)
• Sinusoidial fitting: Fitted curve can be plotted outside data range
• PCO: Plotting on axes 1+3 (and 2+3)
• Rare crash in histogram module fixed

Version 1.76

• Imbrie & Kipp factor analysis of abundance data, with optional regression of environmental data.
• Kruskal-Wallis test now accepts groups with only one value
• Sinusoidal fitting: the fitted constant term is now reported
• UPGMA clustering five times faster
• Individual rarefaction: more options for step size and maximum N for very large data sets

Version 1.77

• Bug fix: incorrect UPGMA clustering (introduced in 1.76)
• Reconstruction of environmental parameters with Imbrie & Kipp transfer functions
• Improvements in semivariogram modelling for kriging
• Reading TPS files: Independent scale tags now supported
• Reading RASC files: Decimal depth format now supported
• Conversion from biostratigraphic FAD/LAD events to occurrence matrix (RASC to UA)

Version 1.78

• 2-way ANOVA bug fix: Incorrect p values when p>0.5
• Polynomial fit bug fix: Incorrect p values when p>0.5
• MANOVA post-hoc pairwise tests bug fix: did not test for p values larger than 1

Version 1.79

• Watson-Williams test for equal mean in two directional samples
• DCA now expects samples in rows and species in columns
• Paste function in Rename Columns
• RASC: Speed increase, and increased maximum number of cycles to 3000
• Bug fix: EDMA crashed for large number of landmarks
• Bug fix: Zero plotting range (giving blank plot) in rare cases

Version 1.80

• Unitary Associations: Speed-up, and addition of several new interpretation modules
• CONOP: Numerical output of ranges
• Bug fix: Kulczynski, Ochiai and Simpson indices were mixed up in the Cluster Analysis module

Version 1.81

• Modern Analog Technique for paleoclimate reconstruction
• Euclidean distances no longer normalized for number of variables
• Individual rarefaction: Multiple samples in one run
• PCA loadings numerical output: All components given
• NMDS now starts without annoying default run
• Wavelet transform: Color option and numerical output
• Bug fix: Crash when reading large Nexus files
• Numerous minor improvements in efficiency, graphics, error reporting.

Version 1.82

• Save graphics in JPG or BMP format (in addition to EMF)
• No. of tick marks can be set independently for both axes
• Mixture analysis: group assignment probabilities for all points
• Modern Analog Technique: semivariogram
• Bug fix: Paried tests crashed for very large data sets
• Graph preferences: graph is now updated instantly

Version 1.83

• All similarity measures available in all relevant windows
• Improved plot of constrained clustering dendrogram
• Improved plot of two-way clustering dendrogram
• Point+line option in Graph and XY Graph
• Copy text from all windows in the Statistics menu
• Significant post-hoc p values marked in pink for ANOVA, Kruskal-Wallis, MANOVA, ANOSIM, NPMANOVA
• Standard font in graphics changed from Small Fonts 7 pt to Arial 8 pt
• "Illegal value in input data" message now reports position of offending cell
• CVA scatter plot is now centered on zero, and "nuisance" axes higher than Ngroups-1 not reported
• Rare crash in NMDS for duplicate points fixed
• A number of other minor improvements in user interface

Version 1.84

• Bug fix in multivariate linear regression: Inaccurate overall MANOVA
• New function: Stack colored rows into columns
• Copy text in polynomial fit
• Fixed occasional 'Fail' for MANOVA post-hoc, introduced in 1.83
• Fixed apparent doubling of topmost row when flipping 'Edit labels'
• Directional stats: New bootstrapped confidence interval on mean
• Normality: Warning for chi2 test when N<20
• Wavelets: Turned upside-down, and added "Power" option
• Detrending now accepts one (y) or two (x and y) columns

Version 1.85

• Bug fix: Occasional crash in Spectral Analysis View Numbers
• Bug fix: Numerical instability in calibration from CABFAC (factor analysis was OK)
• Missing values implemented in chi-squared and Spearman-Kendall (by row deletion)
• Chi-squared: Max total for Monte Carlo test increased to 30000. Improved error messages.
• CCA: All eigenvalues reported, not just four
• New (experimental) plotting method for multivariate time series: "Mantel scalogram"
• Elliptic Fourier Analysis: Number of harmonics increased to 30
• Scripting bug fixes: First row not included when copying data from numerical output window. Function 'fprob' was faulty.
• New scripting functions: abs and anova

Version 1.86

• Skewness and kurtosis now computed as sample values (unbiased estimators)
• Reading Nexus interleaved files: bug fix
• Gridding: Half-cell offset bug fixed
• Bitmap zoom-in bug fixed (matrix, gridding, wavelets etc.).
• Scripting: Progress bar procedure

Version 1.87

• Two-way ANOSIM
• Fligner-Killeen test for equal coefficients of variation
• Univariate statistics: 25 and 75 percentiles were incorrect
• CCA now uses SVD, improving numerical stability for ill-conditioned data
• CCA: copy and print buttons were swapped
• "Point symbols" option in graph preferences (all symbols plot as points)

Version 1.88

• Two-block Partial Least Squares analysis
• Concentration ellipses: Level can be set by user (was fixed at 95%)
• Transform menu: Conversion to row-wise percentages
• Appearence Event Ordination: Numerical output for taxon ranges
• Minor improvements in plotting functions

Version 1.89

• Smoothing spline: New algorithm with weighting, crossvalidation, bootstrapping etc.
• Circular correlation
• Chi-squared Monte Carlo test now supports sampled vs. expected frequencies
• Standardize to unit row vector length
• Transform abundance to presence/absence
• Improved handling of missing data in several transforms
• Evaluate expression: Power with base 0 bug fix, better handling of missing data, new sum function
• Fixed bug introduced in 1.88: Incorrect default graph zoom when including concentration ellipses

• Multiple regression
• One-way ANCOVA
• Measures of beta diversity
• Minor improvements in user interface and error reporting

• Mann-Whitney test now includes exact and Monte Carlo significance for small sample sizes
• Concentration ellipse parameters can be viewed as numbers (from the graph preferences window)
• Individual rarefaction: Curves for all samples can be viewed together
• When pasting into spreadsheet: Only tab is interpreted as separator, not any whitespace
• Missing data supported for Graph and XY Graph (Plot menu)
• Bug fix: Incorrect labels in spindle diagram when selection not from first column
• Bug fix: Crash in log-series abundance model for very large data sets

• Runs test: New Monte Carlo significance value
• Size removal for linear measurements: New method added - removal of allometric size with respect to a single specified variate.
• Parsimony analysis (cladistics): Removed limit of 400 characters. Maximum number of stored trees increased to 10,000. Slight speedup.
• CCA bug fix: Results occasionally corrupted when re-running analysis on same data
• Linear model, 1 independent, n dependents: Numerical output of regressed values
• NMDS: Coefficient of determination given for each axis
• Wavelets: Axis values given in numerical output

• Bug fix: Mixed data distance measure sometimes incorrect
• Bug fix: Occasional crash in permutation test for CCA
• Spindle plot: Option for specifying sample levels
• Mann-Whitney test: Mean ranks given in output
• Diversity t test: Degrees of freedom given in output

• Stereo plot for spherical data, with contouring
• Autoassociation, for sequences of nominal data
• Cross-correlation: Support for missing values
• Walsh transform: More intuitive units on frequency axis
• Spindle diagram: Column labels position bug fixed
• Reading Past files: Faster, and supports Unix-type line breaks
• Elliptic Fourier: Error message for duplicate points
• Taxonomic distinctness: Speed-up
• NMDS: Axis labels not updated when switching from 2+3 to 1+2
• Cluster analysis: Gower distance crash for zero variance, fixed
• Cluster analysis: Gower, Euclidean and Manhattan now shown as positive distances, not negative similarities

• Fixed bug introduced in 1.94: Some row colors incorrectly read from Past files.
• One-way ANOVA now includes omega-squared effect size statistic

• One-way ANOVA: Option for repeated measures (within-subjects)
• Two-way ANOVA: Option for repeated measures
• Discriminant analysis: Visualization of shape deformations along the discriminant axis, both for landmark and outline (EFA) data
• Linear model: Option for robust regression (Least Trimmed Squares)
• Sample rarefaction: Fix to compute variance even if s2=0
• Removed superfluous "Shape PCA" button in PCA
• Box plot and histogram: Crash for some degenerate data, fixed

• SHE diversity analysis
• Kernel density estimation for 2D points
• Monte Carlo version of the Jarque-Bera normality test
• Spline deformation grid for linear regression of landmark shapes
• Diversity profiles: Support for multiple columns, and numerical output
• Normality tests: Support for multiple columns.
• Box plot: Quartiles now computed as in Percentile plot module, with user-selected method
• Added t label for significance of correlation in Linear model
• PCA shape deform: Now allows negative PCA scores, and expansion map
• Landmark linking for 3D: Incorrect labeling of columns, corrected

• Survival analysis: Kaplan-Meier curves; logrank, Wilcoxon and Tarone-Ware tests
• Incorrect reading of some Past files, fixed (bug introduced in 1.94)
• Crash in Mann-Whitney if n2>n1, fixed (bug introduced in 1.91)
• Output of residuals for multivariate linear regression
• Numerical output for kernel density estimation
• Visualization of PCA of Elliptic Fourier Analysis shapes

• Can now open NTSYS files. Trees and multiple tables not supported.
• Risk difference, risk ratio and odds ratio, all with confidence intervals and tests
• LOESS curve smoothing
• Mixture analysis: Now includes exponential and Poisson distributions. Moved to Model menu
• Spearman's rho correlation: Automatically switches to exact test for N<=10
• Ripley's k: Confidence intervals by Monte Carlo instead of approximate equation
• Sinusoidal fitting: Options for fitting to pure sines or cosines (no phase)
• Convex hulls and concentration ellipses can now be color filled
• PLS: Incorrect labels for second block in plot of loadings, fixed
• Linear n indep 1 dep.: Incorrect labels when first column not selected, fixed
• New distance measure: Great-circle distance on Earth's surface (WGS84)
• Bivariate linear regression: r^2 reported in addition to r.
• PLS: "Squared covar %" made available also when using correlation matrix

• Import text file (whitespace, tab and comma separated, label options)
• Numerical inaccuracy in MANOVA/CVA, introduced in 1.83, fixed
• Diversity profiles: Incorrect results for non-integer abundance data, fixed
• Diversity indices: Now accepts non-integer abundance data (bootstrapping still rounds to integers)
• PLS: New equation for "Squared covar %" when used on the covariance matrix (introduced in 1.98)

• One-way ANOVA: Analysis of residuals (plots and tests)
• Visualization of Elliptic Fourier linear regression
• Elliptic Fourier Analysis now also reports sizes
• Smoothing of Mantel scalogram
• Copy text from bivariate linear regression window
• Landmarks: New option for projection to tangent space
• PCA axes automatically flipped if sum of loadings is negative

• REDFIT spectral analysis with red-noise model
• Zero intercept option in linear regression (OLS, RMA and LTS)
• Markov chain analysis for sequences
• Renyi diversity profiles: Now with bootstrapping. Hole removed for alpha=1.
• New algorithm for CVA
• Test for random (Poisson) timing of events
• CABFAC paleoenvironmental reconstruction: Now with k-fold cross validation.
• Multiple regression: Added R^2 for each independent variable
• Polynomial regression: Equation displayed on the plot
• NMDS made much faster

• Linear regression: Major Axis regression included
• Linear regression: Corrected confidence bands for OLS regression with zero intercept (introduced in 2.01)
• Smoothing spline: Option for enforcing monotone curve
• Fitting to logistic equation: Bootstrapped confidence bands. Enforce a=1.
• Survival analysis: Division-by zero for some data sets, fixed.
• Wavelet analysis: Plot of one particular scale
• Mixed similarity measure in NPMANOVA and one-way ANOSIM: Problem with repeated pop-up windows fixed.

• Events-in-time analysis (tests for Poisson process and trend)
• Geographical coordinate transformations (WGS84, UTM etc.)
• CVA: Visualization of landmark and EFA deformations along discriminant axes
• PCA: Plot any axis against any other
• Mixture analysis: Max n=300 limitation removed
• CCA: Percentages of eigenvalues reported
• REDFIT: Monte Carlo simulation now optional
• Correlation: Exact test of Spearman's rs for n<10, not n<11 (was too slow)
• Removed option for Gaussian fit algorithm from species packing modelling (weighted averaging better)

• Directional: Watson's goodness-of-fit test for von Mises distribution.
• Directional: Mardia-Watson-Wheeler test for equal distribution
• Parks-McClellan linear-phase filters (lowpass, highpass, bandpass, bandstop).
• Open files: Support for the FASTA (NCBI specification) molecular sequence format
• Eigenshapes: All axes reported. Fixed rare crash for view scores as numbers.
• Runs test: Options for "runs about the mean" and "runs up and down"
• Modern Analoque Technique: Started crashing in some recent version. Fixed.
• PCO: User similarity and distance indices were swapped in the radio box
• PCA: Grid and EFA deformation now use eigenvalue scaling
• Polynomial and sinusoidal regression: AIC now uses the AIC_chisquared definition with small-sample correction

• Detection of alignments in point patterns
• Quadrat richness: Chao2 updated to latest bias-corrected form. Jackknife1 variance. Bootstrapping.
• Sinusoidal fitting: List of residuals. Occasional phase shift in plot when changing plot type, fixed.
• PCA shape (landmark) and EFA deformation now have option for eigenvalue scaling
• Expansion plots for grid deformation: Visual quality improved, occasional "leakage" of color fixed.
• View numbers in Ripley's K: bug fix in copy function
• Ripley's K: 2x speed-up
• Smoothing spline: Plotting of the given error bars
• Rare crash in Parks-McClellan filter fixed
• Automatic detrending in Spectral analysis
• Autoassociation: P values were not updated when unticking "no repetitions" option.
• PCA: Now reports only rows-1 axes if rows-1 < columns (the rest are zero).
• Correspondence analysis: Plot any axis against any

• Short-time Fourier transform
• Bar chart with error bars
• New distance measure, Mahalanobis, in clustering, PCoA, NMDS, ANOSIM etc.
• CVA now reports classifications and confusion matrix
• Gridding (spatial interpolation) now includes cross-validation
• Semivariogram modeling: Reports residual sum of squares. Automatic model and parameter optimization. New model: Cubic
• Contingency table now includes permutation test
• Fitting to Michaelis-Menten equation
• Expansion map for thin-plate spline deformations: Option for grayscale
• Gridding and 2D kernel density: Option for cropping to convex hull
• Kruskal-Wallis post-hoc analysis by Mann-Whitney now uses the same version of the test as the Mann-Whitney module

• Partial Mantel test
• Cross validation for CVA classifier
• 95% confidence band for LOESS smoothing
• Stereoplots now include density maps (colour or grey scale)
• Read molecular sequence files in the Arlequin format
• Bitmaps made smoother in gridding and kernel density modules
• Procrustes fitting: Option to rescale to original centroid sizes
• MANOVA post-hoc pairwise comparisons: Now reports sequential Bonferroni corrections and Mahalanobis distances. Now based on within-groups covar matrix pooled over all groups, not only the two compared.
• Fitting to growth functions (logistic etc.): Now includes Akaike Information Criterion. New model Gompertz curve. Bootstrapping made more robust.
• Gridding: New gridding method, multiquadric. Cross-validation now reports mean squared error. Kriging made faster.
• Frame around plots.
• Bug fix: Parsimony analysis failed for missing values (bug introduced in 2.06)
• Bug fix: Cluster analysis failed with correlation similarity (bug introduced in 2.06)
• Bug fix: Mahalanobis distance was reported incorrectly in the Two-group Permutation Test module (the p value from the test was correct)
• Bug fix: Shapiro-Wilks normality test gave incorrect results when running several samples with unequal sample sizes (bug introduced in 1.96)
• Bug fix: PCA landmark deformation: Vector plot did not honour eigenvalue scaling option
• Bug fix: Semivariance model "cubic" was not supported by kriging (introduced in 2.06).

• Plot quality improved (antialiasing, floating point coordinates, cardinal splines, alpha blending, etc.)
• Generalized Linear Model, including logistic (logit) regression.
• Partial linear correlation with any number of controlling variables
• Combination of errors (weighted mean, plot of summed probability density)
• Brillouin diversity index
• CCA: option to amplify triplot vectors
• Ternary plot: Show fourth variable with bubble plot or gridding
• New file formats for saving graphic: TIF, GIF, PNG (in addition to JPG, BMP, EMF).
• Wavelets: Plotting of Cone of Influence.
• Linear 1 independent, n dependent regression: Outputs overall MSE and R2. Table of all coefficients.
• 2D kernel density: Choice between four different kernel functions
• Kolmogorov-Smirnov test: Added permutation test.
• NPMANOVA: Option for sequential Bonferroni correction in pairwise post-hoc
• Past file format: Now accepts space character inside cells (double-quoted in file)
• Minor bug fixes.

• Fourier shape analysis by the Haines & Crampton method (Hangle)
• PCA now has three alternative methods for missing data, including iterative imputation
• "Jitter plot" in the box plot module
• Option to flip the axes in Graph, XY graph, XY error bar, histogram, box plot.
• Two-group multivariate permutation test: Added Euclidean distance option
• Bug fix: "Mixed" distance measure in cluster analysis did not handle missing data correctly
• Mixed distance measure: Now robust to failure in one data type due to excessive missing data
• 2D kernel density plot: Row labels and convex hulls
• Bug fix: Histogram for several columns used only first column to set start and end values.
• Concentration ellipses plotted for N>=3 (was N>=2)
• Another bug fix in sequential Bonferroni correction for pairwise MANOVA, NPMANOVA, ANOSIM

• Two-way NPMANOVA
• SIMPER: Added Euclidean, cosine and chord distances
• Beta diversity: Added matrix of pairwise comparisons
• Regular interpolation: New options for point spacing and interpolation method
• Bug fix: Displaced labels in cluster analysis (introduced in 2.09)
• Bug fix: Stratigraphic range confidence intervals: Joint alpha was incorrect for n<=6
• Bug fix: Multiple linear regression: Crash for negative R2 (rare)
• Bug fix: Unitary associations: Crash in biostratigraphic graph
• Bug fix: Concentration ellipses sometimes plotted with very thick lines

• Polyserial correlation (for continuous vs. ordinal variates)
• Multitaper spectral analysis
• NMDS: Optional biplot showing correlation with environmental variables
• One-way ANOSIM: Box plot of within- and across-group distances
• Univariate statistics: Now with bootstrapped confidence intervals
• Survival analysis: Now accepts one or several columns
• All digits retained for all functions in Transform menu
• Fixes of bugs causing rare crashes in NMDS, EDMA, reading of NTSYS files, and compare diversities.

• Friedman test with pairwise Wilcoxon post-hoc
• Spatial autocorrelation (Moran's I) for point data
• Exact version of the Wilcoxon test for n<26
• Bug fix: Wilcoxon test could be inaccurate if any pair contained identical values
• SIMPER now much faster for large data sets
• Short-time Fourier analysis now has option for the multitaper method
• Markov chain analysis can now read multiple columns for pooled analysis

• Multivariate multiple regression (several dependent and independent variables)
• Chi-squared: Fisher's exact test now handles more than two rows (n x 2 tables)
• Linear regression: Durbin-Watson test for autocorrelation of residuals
• Date/time conversions
• Polynomial fit: Table of residuals, and improved error handling
• Short-time Fourier: Bug fix for multitaper method (introduced in 2.12)
• SHE analysis: Improved handling of rare division-by-zero error
• Permutation t test: Speed-up, and max permutations increased to one million
• Matrix plot: Y axis now shows first row at top
• PCA: Fixed rare crash in scatter plot, view numbers
• One-way ANOVA: Copy text now copies whole window
• Contouring: Small improvement in quality

• Simple smoothing filters: Moving average, Gaussian, median, exponential
• Linear regression: Breusch-Pagan test for homoskedasticity of residuals
• Fitting points to logarithmic spiral
• Functions to remove all-zero, all-missing or singleton rows or columns
• Spherical (stereoplot): Calculation and plot of mean direction
• NMDS: Numerical output of correlations with environmental variables
• Pairwise Hotelling's test: Exact test for N<=16.
• F and t test: Bootstrapped confidence interval for difference in means
• Box's M test: Monte Carlo permutation
• Wavelets: New wavelet bases: DOG(m=0) and DOG(m=1). Color scale bar.
• Correspondence analysis: Speed increase for matrices with many more columns than rows
• Similarity and distance indices: Improved error handling
• DCA: Improved error handling. Limit on size of dataset increased.

• Within-group and between-group centered PCA
• Wavelets: Significance test against red-noise hypothesis
• Kruskal-Wallis: p value was inaccurate for data with very many ties
• Bar chart: More options for whiskers (1-sigma or 95%, stdev or stderr)
• SIMPER: Output format made more similar to the PRIMER software
• Linear 1 indep n dep: Now accepts zero stdev on dependents
• CVA: Bug in biplot when only two groups present, fixed
• Crash when viewing phylogenetic trees loaded from Nexus files: Fixed
• Two-group multivariate permutation test now has Euclidean as default distance

• Box-Cox transformation with max likelihood estimation of lambda
• Anderson-Darling normality test
• Nearest neighbor point analysis: Plotting of distance histogram vs. expected distribution
• Hangle (outline shape analysis) now includes 'Htree' fitting
• Coefficient of variation: Now with bootstrapped confidence intervals
• Similarity indices: Added Mixed (as available in other modules)
• Rarefaction: Labels added for multiple-sample plot
• User controllable range of color maps, new color maps
• User controllable number of lines in spline deformation grids
• Bubble plot: Option for 95% confidence ellipses
• Rarefaction: Option to not show standard deviations
• Improvement on number of decimals in graph axes
• Directions: Crash for pathological input, fixed
• Sinusoidal fitting: Crash for pathological input, fixed

• Two-way ANOVA without replication
• Chao-1 species richness estimator (bias corrected)
• Landmark 3D plot: Added animation function
• Coefficient of variation now also available in the Univariate statistics module
• Box-Cox: Automatic addition of constant for zero or negative values
• Kolmogorov-Smirnov: Permutation test was inaccurate, fixed
• Rarefaction: Option to leave out confidence intervals in banana plot
• Cluster analysis: Dendrogram export, displacement of labels bug, fixed
• Wavelet analysis: Color coding changed (again)
• Fourier shape analysis: Number of harmonics increased to 30
• Some other minor improvements

Next: Documentation