Cocorresp Change Log Version 0.4-4 * Suggests: {pls} is now explicitly a suggested package. Version 0.4-3 * Suggests: {rmarkdown} is now explicitly a suggested package following changes to the {knitr} dependencies. Version 0.4-2 * eigenvals: Methods now return an object of class "eigenvals". * print.predcoca: The labels for predictor and response matrices were in the reverse order. The total variances for preditors and responses were correct, but the object names were back-to-front. * print.crossval, print.summary.crossval: Tweak to order of response and predictor matrices in a printed message. Suggested by Cajo ter Braak. Version 0.4-1 * envfit.coca: Don't test the output from example in `envfit.coca` during `R CMD check`. Issue using ATLAS BLAS reported by Brian Ripley. Issue #11 Version 0.3-0 * screeplot: a new `screeplot()` for symmetric Co-CA analyses. * Vignette: a new HTML vignette is provided as an introduction to the co-correspondence analsysi. Version 0.2-3 * Corrected a typo in the package DESCRIPTION. Reported by Prof. Ripley. * Switched to title case in DESCRIPTION Title field. Reported by Uwe Ligges. Version 0.2-2 Development is now done on Github: https://github.com/gavinsimpson/cocorresp * envfit.coca: fixed to work with new vegan permutation interface in `envfit.default`. Reported & fixed by Jari Oksanen * NAMESPACE: import `factorfit` and `vectorfit` as we use these directly. Reported & fixed by Jari Oksanen Version 0.2-1 * envfit.coca: now does weighted correlations in common with `envfit.default` for weighted ordinations. * weights: new methods for objects of classes "symcoca" and "predcoca". * fitted.symcoca: was not preserving the names of the `X` and `Y` community matrices. The `print` method now squishes very small variations in the results resulting from floating point computations. * Tests: The package Examples are now checked against known good runs as part of new package checks in `./tests/`. Version 0.2-0 * NAMESPACE: cocorresp now has a namespace and imports from package vegan's namespace. Hence we now depend on vegan version 1.17-12 or greater. * coca: now returns an object with class "coca" as well as a class from the fitting function ("predcoca" or "symcoca"). * scores.symcoca: Now always returns a nested list even if only a single type of scores is requested. Individual scores components are named consistently `X` and `Y` rather than `U1`, `U2` etc. * scores.predcoca: Now always returns a nested list even if only a single type of scores is requested. Individual scores components are named consistently `X` and `Y` rather than `U1`, `U2` etc. * plot.symcoca, plot.predcoca: Plot method changed to plot only one of the response or predictor results. Now also returns an object of class "ordiplot" which should be able to be used with all of Vegan's ordiplot code. * predcoca.eigen, predcoca.simpls, symcoca: These are now non- visible and not exported. Original intention was that the user would not call these functions. * envfit.coca: a method for the vegan envfit() function for vector and factor fitting external data to Co-CA fits. * eigenvals.coca: a method for the vegan eigenvals() function, which extracts Eigenvalues from Co-CA models. * points.coca: a points method for all types of Co-CA fitted by coca(). * coinertia: this is now a fully user-visible function. The old function is now named fitCoinertia(). Version 0.1-9 * CITATION: Updated citation file. * DESCRIPTION: Updated to latest version - updates were missed from previous version. Version 0.1-8 * CITATION: updated citation file to match CRAN requirements. Version 0.1-7 (opened June 25th 2008) * permutest.coca: Correct markup for S3 method now that permutest is generic in vegan. * pasteCall: This function was added to vegan as a helper. As cocorresp depends on vegan there is no need for it in cocorresp. Version 0.1-6 * Removed Encoding: UTF-8 from package DESCRIPTION file. * Fixed a bug in print method for fitted.predcoca * Added scaleLin and mcLin to internal functions. * Updated Version: field in DESCRIPTION to match new requirements in R 2.6.0. Version 0.1-5 * Added 'stats' and 'graphics' to Depends field. Version 0.1-4 * Completely rewrote plot.predcoca() and plot.symcoca() from the ground up. These methods are far more flexible and modular than before and most options are user-configurable. Methods default to a 1x2 display showing both biplots, setting 'oneFig = FALSE' overides this behaviour and each biplot is plotted in turn on the active device. New argument 'ask' controls whether the user is prompted to display the next plot or not like plot.lm() does. 'ask' tries to work out whether prompting is required (it is not if the user has already set up a 1x2 or greater plotting region). This feature is based on code in plot.lm(). Still need to work out the most intuitive way of handlings set of graphical parameters for the predictor and response for total customisation. * Minor documentation fix Version 0.1-3 * scores.symcoca() incorrectly returned the wrong site scores for matrix X. Reported by S. Juggins. * corAxis.symcoca() incorrectly called scores with argument ''extract'' instead of ''display''. Reported by S. Juggins. * Added flush.console() calls to crossval() and permutest.coca() to make sure the progress is printed to the console in Mac OSx and Windows. Suggested by S. Juggins. * Added argument n.axes to permutest.coca to control the number of axes tested. Defaults to x$n.axes to maintain backwards compatibility. Suggested by S. Juggins. * Minor typos in documentation fixed, and removed \dQuote{} from within \code{} chunks as it was not being processed as I expected. Version 0.1-2 * First Public Release