This page lists the presently known issues with the actual release, followed by the issues fixed since the first release (5.00). Note that the issues reported here include both bugs (corresponding to incorrect software functionality) and features (software functionality that might be implemented in a more useful or easier-to-use way).

How to update to the latest Canoco version

To update your present Canoco 5 installation to the latest maintenance version, select the Help | Check for updates menu command and follow the instructions displayed in the dialog box. Please note that the trial versions, provided by Canoco distributor for evaluating the software cannot be updated.

Issues fixed and improvements added in version 5.15

  • Improvement [515-01]: Added new method of performing Monte Carlo permutation test. Beside the previous method permuting response data (which is also called Y-permutation), users can now choose to permute predictors (X-permutation). Additionally, the algorithm of Y-permutation was mildly improved by adding intercept to covariates, when performing the test with a model including covariates. There is still the option to perform old type ("legacy") Y permutation. The papers describing Y- and X-permutation are in the pdf subfolder of the Canoco installation folder and are called "CCA_permutations_draft.pdf" (most relevant) and "RDA_permutations.pdf" (more theoretical).
  • Improvement [515-02]: Added new statistical summaries to double-constrained ordination analyses. Most (10) of the statistics are provided per individual double-constrained ordination axes, but two adjusted estimates of explained variation are also provided. All these new statistics are shown in the specialised analysis notebook page (called Double Constrained [P]CA), in which the summaries of the dual permutation tests (case-based and response-variable based) and optionally of the tests of dimensionality were placed. The new Copy button allows to copy most of the displayed information (old and new) onto Clipboard. The specialised analysis notebook page also became scrollable.
  • Improvement [515-03]: Added new type of regression biplot to double-constrained ordination analyses, representing the regression of CWM (community-weighted means) on environmental predictors.
  • Fixed bug [515-04]: During the setup of double-constrained ordination analysis, do not allow changing the row weights in analysis steps, to which the weights were transferred from earlier steps. This assures consistency across the steps implementing the dc-CA or dc-PCA method.
  • Improvement [515-05]: Added power transformation (with exponent having the default value 0.1 and allowed range from 0.01 to 4), which can be applied both to a compositional data table (typically with the community composition) or to individual numerical variables in a general data table.
  • Feature [515-06]: Leverage correction of residuals during Monte Carlo permutation test is no longer supported.
  • Fixed bug [515-07]: Made the dc-(P)CA algorithm resistant to producing wrong results when response variable contribution scaling was chosen in a Canoco 5 project. This is now ignored in appropriate steps.
  • Improvement [515-08]: The interactive stepwise selection dialog box now shows in the list of already selected predictor terms the pseudo-F statistic for each predictor at the model selection stage when it was selected.
  • Fixed bug [515-09]: Window graphically summarizing the results of a Monte Carlo permutation test with histogram did not redraw properly when resized.
  • Improvement [515-10]: With limited number of cases, not all constrained axes were calculated. For example, when your constraining (explanatory) variables form a model with 9 degrees of freedom and your dataset has 11 cases, it should be possible to compute all 9 constrained axes (although the dataset is not very good to ask about so complex model), but Canoco estimated only first four.
  • Improvement [515-11]: The list of ordination diagrams offered by Graph Wizard for the results of double-constrained ordination analyses (dc-CA and dc-PCA, including all analysis template variants) was extended, adding in particular the Regression biplot of CWM analysis, which summarizes the (multiple) regression(s) of community-weighted means (CWM) of individual functional traits on the environmental predictors used in the double-constrained ordination. The description of those diagrams - provided by the "Describe contents" command in the Graph menu - was also improved.
  • Fixed bug [515-12]: When modifying existing analysis which specified exclusion of rare species based on their frequency, changes in the set of included cases was not immediately reflected in the set of excluded species and the analysis definition had to be finished and the setup reopened and revised to obtain correct results.
  • Fixed bug [515-13]: When creating an expanded project (i.e. "one response variable per case") using the Project | Create derived project | Expand occurrences... menu command, completely empty cases were not expanded, which does not work well when analysing such data with linear ordination methods.
  • Fixed bug [515-14]: When creating an expanded project, originally empty rows were assigned weights 1.0e-16 rather than exact zeros.
  • Improvement [515-15]: When labelling isolines in contour plots, with fixed number of decimal places specified as preferred format, the trailing zero digits in fractional part should not be removed.
  • Improvement [515-16]: When labelling isolines in contour plots, avoid generation of label for zero value with a negative sign (i.e. labels like "-0.0").
  • Improvement [515-17]: Suggested fixed range of axes for a t-value biplot (-1 to +1) mostly misses the required range for Van Dobben circles to be shown in the graph, when there is just a single constrained axis. Therefore, under such circumstances, the dialog offers and uses (when user keeps the box checked) the range -1.1 to +1.1 instead.
  • Improvement [515-18]: The end of introductory text field, describing the meaning of following information, was not always readable or it was bouncing into following field in the specialized panels of analysis notebooks shown for the analyses based on "compare constrained and unconstrained ordination" or "test individual constrained axes" templates.

 

Issues fixed and improvements added in version 5.12

  • Fixed bug [512-01]: Response variable' cumulative contributions (CContr.* scores) were not calculated correctly if response variable(s) had a missing value for its weight (RespW).
  • Fixed bug [512-02]: Analysis templates for dc-CA and dc-PCA method including stepwise selection of traits and/or explanatory variables did not carry over correctly a chosen factor due to automatic selection of its last (collinear) level.
  • Fixed bug [512-03]: Contribution-based response variable scores (introduced in version 5.11) were not correctly calculated if response variable(s) had a missing value for its weight (RespW).
  • Feature [512-04]: dbMEM scores calculated with weighted averaging used case weights in the principal coordinates analysis of the adjusted distance matrix. In this way, dbMEM scores calculated from an analysis using linear method (typically partial RDA) differed from those in unimodal methods. While this was not necessarily a wrong approach, it does not adhere to the original description of the method.
  • Fixed bug [512-05]: Community-weighted means (CWM) data table was not calculated correctly with the Data | Add new table(s) | Trait averages... menu command, when there was a factor trait with missing values.
  • Feature [512-06]: Rao's measures of functional diversity did not enforce the distance values into 0-1 range, so depending on the chosen type of trait standardization, the relation to Simpson D could be lost and hence the transformation of FD to the scale of effective number of species could be incorrect. Now the calculated distances among response variables (taxa) are always standardized to (0,1) range except when the traits are pre-standardized to (0,1) range.
  • Improvement [512-07]: The definition of response variable ("species") groups by a rule was extended to allow their selection based on the CContr.* (cumulative contribution to axes definition) scores; at the same time, a choice to define group of cases based on their heterogeneity (Het) scores was removed, as not frequently used. They can be still used for group definition by copying the required scores to a data table before defining the group.
  • Fixed bug [512-08]: Dual diagrams of CoCA did not remember the choice of the plotted response variables (based on their CFit) from either of the two compared compositional data table. So upon re-creation, all variables were included. Similarly, ordination diagrams from DCA (or DCCA) did not remember the choice of response variables, based on their weight (CFit is not available when detrending by segments).
  • Improvement [512-09]: Calculated indicator values of response variables of compositional data tables are now accompanied by the estimates of their significance based on an unconstrained permutation of class membership, as suggested by Dufrene & Legendre (1997). This is provided also for the alternative, newer indicator value coefficients of Phi type. Those p-values can be optionally adjusted using the false discovery rate approach.
  • Improvement [512-10]: Few help topics (particularly the one describing the options for calculating a new data table with case-related statistics of a compositional table) were improved.
  • Improvement [512-11]: Graph-wizard offered diagrams for dc-CA and dc-PCA analysis templates (including those with stepwise selection of predictor variables) use compromise scaling of ordination scores, rather than the default scaling set for unimodal or linear methods. This gives a better visual presentation particularly if the case-constraining predictors (e.g. environmental variables) are plotted together with response-variable-constraining predictors (e.g. species traits). If such a diagram is re-created, however, Canoco reverts to the presently setup default scaling.

 

Issues fixed and improvements added in version 5.11

  • Feature [511-01]: PDF files with the most important changes in the new printed manual (distributed since Canoco version 5.10) are now available to those who simply updated their version 5.04 or older, in the pdf/ManualChanges subfolder of the Canoco installation folder. The easiest way how to access them is by opening the manual-changes.html file, located in the pdf folder.
  • Improvement [511-02]: Improved algorithm for computing cumulative fit (CFit) values in dc-CA and dc-PCA methods.
  • Feature [511-03]: When importing weights for cases or response variables in the Analysis Setup wizard with some of the weight values too extreme (so they must be adjusted), the report is now more precise about the changes. Also the related help topic was improved.
  • Fixed bug [511-04]: The helper used in some analysis templates (namely those with "FS" in their names) for transferring the choices of stepwise selection from one analytical step to another was sometimes working incorrectly, namely if a variable was automatically excluded due to its collinearity with selected predictors.
  • Improvement [511-05]: If in an Analysis Setup wizard session the predictors need to be selected into multiple groups (e.g. in the variation partitioning templates or in partial-constrained ordination), their selection can be now based on an existing predictor group (similarly to how response variables or cases can be selected).
  • Improvement [511-06]: When creating a t-value biplot for an analysis with just a single predictor, the dialog box with plotting options is now also shown (it was skipped in the previous versions). In addition, the option for the range of axes to be restricted to -1,+1 is now pre-checked, as it is an imminently useful default choice.
  • Fixed bug [511-07]: Chessel biplot scores for traits (BipChT scores) could be wrongly calculated (with a wrong sign) in the case an axis direction reversal had to be applied (the sign change was then applied twice).
  • Improvement [511-08]: The error message box shown when trying to open a Canoco 5 project with too old (outdated) version of Canoco is now more helpful in suggesting how to fix the problem (i.e. by updating the Canoco installation).
  • Improvement [511-09]: Response variable (Resp) scores are now offered in alternative contribution scaling, both for unimodal and linear methods (in linear methods, contribution scaling is incompatible with the post-standardization of Resp scores by the standard deviations of response variables, so these two choices are mutually exclusive). The new choice is available from the Scaling Options dialog box, accessible from the Canoco 5 toolbar. Additional information can be found in the ContributionsInCanoco.pdf document in the pdf subfolder of the Canoco 5 installation folder.
  • Improvement [511-10]: Irrespectively of chosen scaling for response variable scores (see the previous point), Canoco 5 now calculates cumulative contribution (CContr) statistics for individual response variables and they can be used, alternatively to response variable weights, when selecting the variables appearing in the ordination diagrams (as accessible from the second page of the dialog box shown by the Analysis | Plot creation options menu command). Additional information can be found in the ContributionsInCanoco.pdf document in the pdf subfolder of the Canoco 5 installation folder.
  • Fixed bug [511-11]: When defining groups of response variables (or cases), rule-based selection using their scores on ordination axes did not properly apply the current score scaling.
  • Improvement [511-12]: The "Describe contents" utility for ordination diagrams was completely revised, with a new wording for the suggested interpretation, including more context-specific warnings (based on the ratio of eigenvalues of the displayed ordination axes) about sub-optimal scaling choice for a particular interpretation. The descriptions reflect correctly the various chosen types of scaling and all the possible combination of items, with separate texts provided also for regression biplots, t-value biplots, canonical variate analysis plots, or principal response curves diagrams.
  • Feature [511-13]: When plotting the results of a detrended correspondence analysis (DCA), the Graph Wizard offers a limitation of the number of plotted response variables (typically "species") based on the limit referring to CFit values, despite in DCA, species weights are used.
  • Improvement [511-14]: The co-correspondence analysis (CoCA) now calculates approximate cumulative fit (CFit) values for variables in both compared data tables (labelled as CFit2 scores for the second data table). These scores can be used to select variables shown in the dual plots of CoCA.
  • Improvement [511-15]: Groups and classifications defined in a particular project can be now copied to Clipboard and pasted into another project. The only requirement at the receiving side is that the number of items of chosen type is identical as in the source project. Rule-based definitions of groups and classifications are automatically turned into manually-selected ones.
  • Feature [511-16]: When selecting response variables to use in an analysis based on their frequency (number of non-zero values), the Analysis Setup wizard used their frequency in the original (complete) data table. As a new feature, the frequency is now calculated only from a subset of cases left in the analysis - the cases excluded due to missing values or explicitly removed by the user are not considered.
  • Fixed bug [511-17]: When plotting diagrams with case scores in distance-based analyses (PCoA, NMDS), requested subsetting of plotted cases (N cases with the best fit) was neither respected, nor remembered.

 

Issues fixed and improvements added in version 5.10

  • Improvement [510-01]: When computing community-weighted means (CWM) of trait data, allow for pre-standardization of trait values to unit variance, including also an option for a weighted unit variance
  • Fixed bug [510-02]: changes in transformation during analysis setup, caused by clicking the "Re-advise" button, were not transferred properly to following steps, producing inconsitent results in multistep analysis, e.g. in variation partitioning
  • Improvement [510-03]: added a "Transpose compositional table" command into the Data | Add new table(s)submenu. Identity of new table' rows is that of the source table' columns. Transposed community table is needed to use the new double-constrained CA and double-constrained PCA templates.
  • Improvement [510-04]: added multiple analysis templates for the newly supported double-constrained CA (dc-CA) method. These templates are available whenever user submits four data table to New Analysis Wizard: community table, transposed community table, environmental data table, and species trait data table. One template expects the user to choose proper sets of environmental variables and functional traits to use, while the other offers interactive stepwise selection from both sets of predictors. Both templates also perform appropriate two-stage testing, combining tests based on CWM scores and on species responses to environmental variables. Additionally, two templates were added for testing individual constrained axes of dc-CA, one with the case-based permutation test, the other with the response-variable based permutation tests.
  • Improvement [510-05]: implemented a linear counterpart to dc-CA, namely dc-PCA, again with the same four versions (including stepwise selection of predictors from the two sets) as in the case of dc-CA.
  • Fixed bug [510-06]: use local decimal point separator when exporting results onto Clipboard or into text format files, as well as local list separator when exporting into files with CSV format. This new behaviour can be overriden, though, with a new option "Export data with local separator" at the Data Sheets | Actions page of the dialog box shown by the Edit | Settings | Canoco5 options menu command.
  • Fixed bug [510-07]: non-zero values were generated from zero values or missing (NA) values during the extended log(A*x+B) transformation, overflowing data storage for response data.
  • Improvement [510-08]: enabled stepwise selection of explanatory variables and overview of simple and conditional effects for a CCA method with active detrending.
  • Fixed bug [510-09]: when calculating NMDS or PCO (or db-RDA) on a compositional data table with preset centering/standardization by rows or columns, these operations were (incorrectly) applied to data before calculating distance matrix.
  • Improvement [510-10]: added Shepard diagram to NMDS analysis template, which is now shown (together with STRESS criterion value) in a dialog shown by the Details button in Summary page of the analysis notebook for the first (NMDS) step. The graph can be copied to the Clipboard.
  • Improvement [510-11]: Canoco 5 became A LITTLE more obtrusive towards its users and now checks on its start and it will inform *once* the user about new version availability. This would be done, however, for every separate user account on the computer where C5 is installed. If user skips a version (e.g. stays with 5.10 even after version 5.11 became available), the update about the existence of a new version is repeated for the following ones (e.g. when version 5.12 becomes available).
  • Improvement [510-12]: the corrections of significance (P) values now uses more appropriate penalty when testing, during stepwise selection or evaluation of simple and conditional effects, a factor, where only k-1 sensible tests are done for a factor with k levels. This reduction of penalty also extends to analysis-defined interaction terms, at least the simple ones.
  • Improvement [510-13]: percentages of explained variation for individual explanatory variables are now shown in the summary of simple and conditional effects with adaptive display precision (depending on how many small values are present) - with one to three decimal places for the percentage (0-100) scale.
  • Feature [510-14]: fill style can be now set for enclosing envelopes when editing global visual attributes (using the Edit | Settings | Visual attributes command).
  • Improvement [510-15]: added possibility to create - in a data table - new numeric variables representing together an interaction between one factor and one numeric variable. The new "Create numeric-factor interaction" command is available both from the Edit menu and from the context-sensitive popup menu.
  • Improvement [510-16]: Canoco 5.10 now warns with a (suppressable) message box when cases are omitted from the analysis due to their emptiness (CA/DCA/CCA methods but also some distance-based analysis cases).
  • Improvement [510-17]: added command to double-center (horizontally as well as vertically) selected label(s) in a graph. This is now offered both in context menu and in the Edit menu.
  • Improvement [510-18]: default export format is now CSV when exporting a data table.
  • Fixed bug [510-19]: CSV and TSV export now always presents missing value as "NA" and zero values as "0", rather than as an empty string.
  • Improvement [510-20]: Canoco 5.10 supports import of condensed data in a text format (CSV or TSV). Ordering and identity of rows and columns can be derived either from file codes (if they are integer values) or from optional files coding for row and/or column labels. Example files can be found in the Samples\ImportCondensedData folder.
  • Fixed bug [510-21]: the dialog box offering statistics to compute from a compositional data table did not update its offer when changing to another data table (if more than one was available within a project)
  • Improvement [510-22]: added alternative way to calculate "enclosing" ellipses for classes of objects; the new ones approximate 2D-Normal spread of positions, based on variance-covariance of X and Y coordinates, and display for the estimated bivariate Gaussian distribution either 95% or 66% confidence areas.
  • Improvement [510-23]: added two new analysis templates for Aitchison's weighted log-ratio PCA and its constrained form (log-ratio RDA)
  • Fixed bug [510-24]: tooltips for the choice of permutation test on the first constrained axis vs. on all constrained axes were swapped in the corresponding page of analysis setup wizard.
  • Improvement [510-25]: Canoco 5 adviser now gives more specific advices (in the ordination method page of the Analysis Setup Wizard) concerning the choice of ordination method for the usual unconstrained and constrained analyses. Most of changes happen for the case where - for a compositional response data table - test DCA can be calculated and the lengths of ordination axes (in SD units) estimated (but there is also an extension of advice for purely binary data, where the linear method is mentioned as an alternative). For intermediate and short gradients, the advice also depends on the percentage of zeros. For the case of short gradient and very few zero values, log-ratio analysis is suggested as a viable alternative to a normal linear model (and unimodal method is not mentioned).
  • Improvement [510-26]: Canoco 5 adviser now informs during analysis setup (in the textual advice about ordination model or when leaving the page with ordination method options, depending on method type) about the impact of chosen method (unimodal) or chosen standardization by cases (linear methods) on the existing empty cases (i.e. just with zero values in the response data).
  • Fixed bug [510-27]: Predicted or fitted response variable values (via the Data | Add new table(s) | Predict from constrained ordination) were not precise. Now it gives a correct fit and predicts exact values when the amount of explained variation is 100%.
  • Improvement [510-27]: New method implemented is PRC (principal response curves) based on CCA, i.e. on a weighted averaging algorithm. This is implemented again as two templates, one focusing on the first set of principal response curves, the other allowing to test and display up to four such sets, corresponding to the first four axes of partial CCA
  • Improvement [510-28]: The two sets of analysis templates focused on relating species attributes (aka functional traits) to plot attributes (aka environmental descriptors) were reworked. First one is now called Double Constrained Ordinations and contain the eight templates implementing (with various specific tasks) the two new methods in Canoco 5.10 - dc-CA and dc-PCA. The other group (set) of templates is now called Interpreting by Traits and the five analysis templates present in it uses trait data table as explanatory or supplementary variables in analyses, in which as response variables serve the species scores from analysis (analyses) either unconstrained or constrained by environmental predictors. Analyses from both groups of templates are illustrated by the revised example project Traits.c5p. All analyses now use just a simple approach to focal data table: community data table is always the focal one.
  • Improvement [510-29]: We have changed the way the pseudo-F value is calculated for the test on first constrained axis. This corresponds to a change in formula (3.13), where lambda-1 (numerator) is now divided by q, the number of DFs for explanatory variables. This makes new pseudo-F values for a test on first axis look usually smaller than they used to be, but no effect on p-values and their scale is now compatible with the pseudo-F values for the test on all constrained axes.
  • Fixed bug [510-30]: Canoco engine no longer reports "Axis is null, so cannot be turned into covariate" in the scenario where we ask for calculating too many axes (based on the count of explanatory variables), but the count of axes with non-zero eigenvalues is much more restricted by the number of response variables (this happens particularly in the trait-related analyses where the response variables are Resp.x scores from a previous analytical step).
  • Improvement [510-31]: Canoco no longer asks the user whether to store project' data tables in a condensed form. This is now decided internally.
  • Fixed bug [510-32]: Canoco was copying scores and/or data tables to Clipboard an excesivelly long time when there were many rows of them (> ca 10 000).

 

Issues fixed and improvements added in version 5.04

  • Improvement [504-01]: Improved Canoco 5 performance by parallelizing the bottle-neck computations on multi-core computers using OpenMP library.
  • Improvement [504-02]: Switching between brief and non-brief versions of analysis notebooks was made more easy, by adding toolbar button switching the brief display state on and off for each notebook separately.
  • Improvement [504-03]: Canoco 5 now adapts more readily to user's idea about the required number of permutations: specifying new value in the Analysis Setup Wizard leads to it being accepted as a new default in the analyses defined afterwards.
  • Fixed bug [504-04]: Sorting response variables in the score page of an analysis notebook did not always work in linear ordination methods. This is because the sorting order was based on the default scaling option, but division by response variable' standard deviation could change the order.
  • Improvement [504-05]: Increased the number of case (and response variable) classes and improved their visual consistency (e.g. series line should use same colour as used by the corresponding class symbols, the types of symbols used for cases shall not overlap (much) with the types used for response variables or explanatory variables, etc.)
  • Improvement [504-06]: The cases made supplementary during the analysis execution (e.g. because the explanatory variable or supplementary variable values are missing) are no longer permanently marked as supplementary in the analysis settings. This makes easy - after replacement of missing values - to reexecute the analysis with all the cases active.
  • Fixed bug [504-07]: The Help button in the Define Grouping dialog box (invoked from the role pages of the Analysis Setup Wizard) did not display corresponding help topic.
  • Improvement [504-08]: The Create New Analysis wizard was too small to allow seeing properly the offered analysis templates in its last page.
  • Improvement [504-09]: The error message shown when the TAKE/SKIP rule is incorrectly specified in the Analysis Setup Wizard was confusing (about the number of expected units). Improved the message and the corresponding help topic.
  • Improvement [504-10]: In the db-RDA analysis templates, user choice of omitted (or weighted) cases made in the first step (performing PCO) are now carried over to the second step (performing RDA with PCO scores).
  • Feature [504-11]: The Compare-constrained-unconstrained analysis templates (optionally with covariates) matched CCA with CA method, instead of DCA and detrending could not be even set.
  • Fixed bug [504-12]: The dialog box shown during stepwise selection when the R2(adj) of selected predictors is about to exceed the R2(adj) of the whole pool of predictors could interrupt the automatic selection of all levels of a factor.
  • Fixed bug [504-13]: Cases with missing values in explanatory variables were not detected if these variables are represented in the model just by their interaction.
  • Fixed bug [504-14]: Canoco 5 crashed on GAM models where one of the predictors has just three distinct values.
  • Improvement [504-15]: Allow the choice of Citation button in the About dialog box using keyboard short-cut, Alt-C.
  • Fixed bug [504-16]: The dialog box shown by the Graph | Attribute plots | In ordination space did not allow explanatory variable for plotting when there is just one, producing single constrained axis.
  • Fixed bug [504-17]: The labelling of levels for a nominal explanatory variable in the co-inertia analysis was wrong (level labels are mismatched).
  • Improvement [504-18]: Made the choice of analysis templates in the last page of the New Analysis Wizard easier by adding +/- buttons allowing to open/close a group of templates with a single click, instead of double-clicking group name.
  • Improvement [504-19]: When user closes Log pane of a project, this is now remembered when opening the project file in a new session.
  • Fixed bug [504-20]: Case symbols in ordination diagrams of db-RDA analyses, as offered by Graph Wizard, now represent the CaseE, not CaseR scores for constrained axes.
  • Fixed bug [504-21]: When re-executing already defined analysis, the execution failure led to a complete restoration of the original analysis state, so that no error messages, explaining the reasons for failure could be found in the analysis log.
  • Improvement [504-22]: Changed Canoco 5 so that it supports simple topology-based eigenvector maps (sensu Griffith & Peres-Neto 2006, Ecology) in spatial analyses. Square connectivity matrices (with 0 for disconnected sites and 1 for connected ones) can be specified as similarity matrices (with 1-S transformation to distances) in PCNM analyses, but only if the PCNM threshold value is specified to be computed so that (N-1) nearest neighbours are covered, where N is the total number of sites.
  • Improvement [504-23]: Support optionally the "permutational form of adjustment" of the explained variation in constrained analyses (including variation partitioning), as suggested by Peres-Neto et al. (2006, Ecology). This is based on estimating the amount of explained variation when the explanatory variables are randomised (during Monte Carlo permutation test).
  • Fixed bug [504-24]: when creating a copy of existing analysis (using Modify button) with Hellinger standardisation specified for cases, this option was not copied.
  • Fixed bug [504-25]: the "CopyVarSelection" helper was not properly replicated when creating a copy of analysis (e.g. the PCNM-FS-covariates type).
  • > Fixed bug [504-26]: When executing menu commands affecting data table contents, cell being presently edited was not correctly changed and further it was locked into editing mode.
  • Fixed bug [504-27]: After changing font settings for data table editor, column labels disappeared during scrolling.
  • Fixed bug [504-28]: Popup menu for graph objects was not correctly positioned after horizontally splitting analysis notebook.
  • Improvement [504-29]: Many help topics needed language/contents improvement.

 

Issues fixed and improvements added in version 5.03

  • Fixed Bug [503-01]: The 32-bit version of Canoco 5 crashed when executing (D)CA on a data set with 'perfect split' of the data table into two disjunct sub-tables, i.e. when the first eigenvalue is >= 0.999.
  • Improvement [503-02]: Allow custom aggregation of cases of a compositional data table (via the Project | Create derived project | Aggregate cases). Instead of averaging, data tables (particularly the compositional ones) can be aggregated using sum, median, variance or inter-quartile range.
  • Improvement [503-03]: Added possibility to "unlock" the analyses created by Canoco Adviser, so that they can be further customized (using new Clone unlocked command in the Analysis menu). This is useful particularly if you want either change the type of permutation testing in advanced analysis templates (e.g. to calculate simple and conditional effects in discriminant analysis) or when you want to replace existing data table by another one. This is an advanced feature not recommended for less experienced Canoco 5 users.
  • Fixed Feature [503-04]: Improved user apprehension of program interface when specifying interaction terms in analysis templates using the same data table for explanatory variables and covariates. These two are specified on separate setup wizard pages, shown in non-QuickWizard mode, and when a user checked for interaction definition for explanatory variables at the earlier page, (s)he might have been surprised to see, on the later page, that the definition of interactions was pre-checked also for covariates. Unchecking this box, however, switched off the interaction definition for explanatory variables as well. The fix was to disable the checked box at the later page and allow the change of state only through the explanatory variables area, in this particular context.
  • Improvement [503-05]: Changed wording of the Do not draw symbols on the curves for the specialized page of Graph Wizard used for principal response curves (PRC) diagram. The new wording is positive (Draw symbols on the curves) and the option meaning (and default value) is therefore flipped.
  • Improvement [503-06]: Changed wording of the second page of the New analysis wizard ("Select table focal for"... to "Select focal table for"...).
  • Improvement [503-07]: Better support for automated labelling of ordination axes with the ordination method name and axis number: when user asks for the dialog for axis labelling and transformation to be shown each time a diagram is created, these axes labels are not only be generated (as they already were) but also selected for plotting by default. Note that this was already the case for XY(Z) diagrams.
  • Improvement [503-08]: Made the second page of Excel import wizard more clever by suggesting the number of imported data tables, depending on how many source sheets were selected.
  • improvement [503-09]: Added support for silent installer (and uninstaller) execution.
  • Fixed Bug [503-10]: Few of the registry data (concerning the automated opening of .c5p and .c5g files in Canoco 5 application) were left uninstalled on 32-bit Windows platform when Canoco 5 package was unistalled.
  • Fixed Bug [503-11]: Fixed error in discriminant analysis (CVA) template, where the Setup Wizard, when executed in non-QuickWizard mode, did not offer the selection of response variables and hence all the variables present in the table with case classification factor were used, leading to an incorrect analysis.
  • Fixed Feature [503-12]: The Downweighting rare <response-variables> option in the Analysis Setup Wizard had no tooltip shown.
  • Fixed Feature [503-13]: When right-clicking a centroid (or, less usually, an arrow) of explanatory (or supplementary) factor variable in an ordination diagram, creation of attribute plots is no longer offered. While this submenu was originally offered, no attribute plot was, in fact, created for such non-numeric variables.
  • Improvement [503-14]: Made use of Excel import wizard easier in a situation, where second or later imported data table is transposed (90 degree rotated) relatively to the first data table, and the user forgets to check the corresponding option before entering the spreadsheet area preview. This context often led to difficulties, as the user tried to get back (to check the corresponding option on the preceding page), but the wizard prevented the return based on its check of matching row numbers. Now, only the progress in forward direction from the spreadsheet preview page is prevented on row mismatch.
  • Fixed Feature [503-15]: Canoco Adviser was too hesitant in offering suggestions for response data transformation in circumstances where the type of ordination method (linear vs. unimodal) was preset in the analysis template (e.g. for PRC analysis) or it was implied by data nature (e.g. only a single response variable entering the analysis). This resulted, e.g., in a situation where the log transformation was never recommended for PRC analysis, even where this was appropriate, given data properties.
  • Fixed Feature [503-16]: CoCA analysis template with specified supplementary variables claimed in the Summary page of the analysis notebook that these variables explain 0% of the variation in the CoCA ordination space, but this was due to the fact this statistic is not computed. Changed the way of reporting.
  • Fixed Bug [503-17]: When an attribute plot was created with attribute values coded only by a colour gradient (colour ramp), the request to label the symbols by attribute values was not respected.
  • Improvement [503-18]: Provided better visual feedback when complex diagrams are re-created or re-drawn. An hourglass cursor is now shown also when re-creating or re-drawing an existing diagram.
  • Fixed Feature [503-19]: When importing from an Excel worksheet multiple tables and their case labels disagreed and they were imported as full labels, Canoco 5 incorrectly reported the total number of labels being imported as 0,in a message box.
  • Improvement [503-20]: Allowed silent execution also for a binary patcher of older Canoco 5 versions.
  • Improvement [503-21]: Binary patcher for older Canoco 5 versions made more robust in finding the folder with installed Canoco 5.
  • Fixed Bug [503-22]: Fixed Canoco crashing when one of the envelopes could not be computed and also made the envelope calculation more robust.
  • Fixed Bug [503-23]: Contours in attribute plots were not always properly scaled, namely when combined with different ordination items and therefore forced to be re-scaled.
  • Improvement [503-24]: Users can now specify that the legend section layout wraps after more than four sections.

 

Issues fixed and improvements added in version 5.02

  • Bug [502-1]: Change of label text could not be undone and did not raise the "project changed" state.
  • Feature [502-02]: The decision which predictors (explanatory/supplementary variables) have their centroid scores (CenE/CenS) computed was based on positive value of their average and for those variables where calculation was not done, zero value was returned, leading to symbols presence in graphs. Now are the centroid scores computed for all predictors with no negative values and when not calculated, missing (NA) values are used.
  • Improvement [502-03]: Within Analysis Setup Wizard, pre-defined group of items (e.g. cases or response variables) can be used to quickly select a set of items used or excluded from an analysis or a set of items to be weighted.
  • Improvement [502-04]: Created a simple utility (C5CleanReg) that removes user settings for Canoco 5 from the Windows registry. This utility is provided mainly for user support, but can be downloaded from here.
  • Improvement [502-05]: Removing typo in the message informing user about available Canoco 5 update (redundant closing parenthesis character).
  • Improvement [502-06]: Progress bar shown during analysis execution now progresses more smoothly, being updated during permutation test (in each analysis step, for multi-step analyses).
  • Bug [502-07]: For some of the example projects (e.g. Grid.c5p project in Samples/Permutio folder), limiting number of response variables shown in graph had no effect (both when specified in Graph Wizard or by Analysis | Plot settings dialog box before creating the graph).
  • Improvement [502-08]: AIC calculation changed: Canoco 5.02 no longer uses the "AIC-proportional" statistic, but rather the true statistic value. The finite-sample correction is also applied (calculating AICc values) by default, but this can be switched off.
  • Bug [502-09]: There was a problem with permutation test settings propagating to new analyses, when in a partial constrained ordination, covariate Z was used to define permutation blocks and then a new variation partitioning analysis was created, where one of the groups contained variable Z.
  • Improvement [502-10]: When user chooses to limit the set of plotted response variables to N best fitting ones in Graph Wizard, this choice is propagated to analysis' plot settings (unless the user already selected a different value there). This has the nice consequence (in most types of use) that the number of plotted variables does not increase after recreating the diagram created with Graph Wizard.
  • Improvement [502-11]: There is a new scroll-button above the upper right corner of the list of analyses, allowing to move currently selected analysis upwards or downwards in the list. This complements the existing possibilities of renaming or deleting an analyses, so that the user has a complete control over the arrangement of analyses in the project. Change of analysis order is persistent (stored in the project file).
  • Improvement [502-12]: All submenus of the main menu can be opened by a keyboard shortcut. In the previous versions, this was not possible for Data and Analysis submenus, where Alt-D and Alt-A shortcuts applied to the controls of the project pane. With this improvement, one can easily characterize the required menu command choice, such as "Alt-A P" for opening Analysis | Plot creation options dialog.
  • Improvement [502-13]: In the dialog for the response curves command (in the Graph | Attribute plots submenu), the response variables available for plotting can be either kept in their table order or sorted alphabetically, for an easier lookup.
  • Improvement [502-14]: Improved the wording on specialized setup wizard pages shown when working with distances (NMDS, PCO, PCNM, db-RDA analyses): the direct input of distances from an existing table is now described in less confusing way
  • Bug [502-15]: Canoco 5 stopped redrawing user interface (analysis notebook contents and usually also the main menu) when more than ca 30 analysis notebooks were open in a project.
  • Improvement [502-16]: The action performed by the "Add table..." button in the project pane was changed: it no longer offers creation of a new empty data table, but rather adding one or multiple data tables via import from Excel, which is a more frequented path for Canoco 5 users.
  • Improvement [502-17]: The Log page of each analysis notebook now displays the list of covariates used in each analysis step. Note that for factors, the active covariate levels are shown in the list.
  • Improvement [502-18]: The analysis formula shown in the third page of the New Analysis Wizard now makes obvious that a priori covariates can be added to PRC analyses (with QuickWizard mode switched off). This information was also added to topics describing the PRC analysis templates.
  • Improvement [502-19]: The execution of an analyses with more than 100 explanatory or supplementary variables was substantially speeded by omitting the calculation of the matrix of correlations among individual predictors. The matrix calculation (and storage in the Log page) can be still switched on using an option in the Edit | Settings | Canoco5 Options dialog box, on its Analyses / Actions page.
  • Improvement [502-20]: The main (left-hand) plot in the specialized PRC diagram now labels the horizontal axis with "Time" rather than the less clear "PRC.Time".
  • Improvement [502-21]: The axes of an XY diagram (or of one-dimensional X-diagram or Y-diagram) referring to ordination scores now respect the "Flip axes" options in the dialog shown by the Analysis | Plot creation options menu command. Note that the choices are interpreted with respect to horizontal vs. vertical direction, not with respect to the order of currently active ordination axes. Practical implication of this new feature is that the vertical axis of PRC diagram (and of its supplementary plot with response variable scores) can be now reverted. On the other hand, the horizontal (time) axis of the PRC diagram is protected against the reversal.
  • Improvement [502-22]: Implementation of the "average of table weights" choice in the specialized setup wizard page of co-correspondence (CoCA) analysis template has changed. Now the relative case weights are averaged, rather than the original row totals that were dependent on the average value magnitude and number of response variables in both compared data tables.
  • Improvement [502-23]: When importing factor variables from Excel data sources, a warning message is shown. When this happens for non-compositional data tables, this is rather an informative message box and its display can be suppressed by the user. For compositional data tables, however, the user is advised that the imported factors should be (likely) transformed to numeric values before analysing the data.
  • Improvement [502-24]: Added a new analysis template for a partial unconstrained analysis (PCA or CA or DCA with covariates).
  • Improvement [502-25]: Implementation of PCNM spatial eigenvectors calculation is now up-to-date with the recommendations of the Dray et al. (2006) paper in Ecological Modelling, namely in (a) zero distances on the main diagonal are now replaced with the 4 * threshold value, so that the points are not auto-correlated with themselves; (b) the spatial eigenvectors corresponding to negative eigenvalues can be not only computed but also used in the PCNM analysis template to model spatial autocorrelation with negative Moran's I values. Additionally, the algorithm for calculating the adjusted distance matrix for PCNM eigenvectors was improved so that it now handles correctly simple connectivity matrices (with 0 and 1 values), even when they are non-symmetrical.
  • Improvement [502-26]: Another improvement for PCNM method is the addition of a new analysis template where a priori covariates are added to PCNM variation partitioning, so that for example a global spatial trend (defined by X and Y coordinates and optionally their polynomial terms) can be removed from the response data.
  • Improvement [502-27]: Further change to PCNM method is the suppression of the global permutation test on PCNM eigenvector pool before their stepwise selection. This suppression takes place when the corresponding constrained ordination model is too much determined, with the number of eigenvectors approaching the number of cases, or when their count is generally large (over 100).

Issues fixed in version 5.01

  • Feature [501-01]: to display classified symbols in a XY diagram, one had to check the Use inclusion rules in the Create XY(Z) Plot dialog.
  • Bug [501-02]: after selecting the Compare-constrained-unconstrained-* analysis template, the Canoco Adviser might recommend using unimodal ordination method. If user rejected this suggestion and changed to linear method, the unconstrained method of the second step was correctly changed, but the detrending option was not switched off. Consequently, the helper suggesting calculation of more than 4 axes (so that all constrained axes are computed) did not work and displayed message box.
  • Bug [501-03]: when displaying an ordination diagram with the scatter of explanatory / supplementary variables including both arrows and centroids, for a DCA or DCCA with detrending by segments, and there were neither case nor response variable scores in the diagram, the centroids and arrows were mis-aligned, as the arrows were not shifted by the values of coordinate origin.
  • Bug [501-04]: when you have defined in Analysis Setup Wizard a third order interaction (e.g. A:B:C) with a help of previously defined second-order interaction (e.g. B:C combined with A) and then asked for second-order interaction removal, this did not happen: the Canoco engine used both B:C and A:B:C interaction terms.
  • Bug [501-05]: model DFs used in the warning about model DFs exceeding total DFs were calculated incorrectly, using number-of-levels as DFs for each factor, without decreasing their count by one. The interactions among factors were affected in the same way.
  • Bug [501-06]: after creating attribute plot with contour lines, the restriction of axes range so that not all original contour lines are included lead to contour lines being drawn outside the axes rectangle.
  • Feature [501-07]: Canoco engine did not compute CenE/CenS scores for explanatory / supplementary variables with average (weighted average for unimodal methods) smaller or equal to 0.001. Such a case could happen for example, when cases belonging to particular category represented less than 0.1% of the total number of cases (or even more, if such cases have below-average weight in unimodal methods). Now the threshold is at 0.000001.
  • Bug [501-08]: when exporting a graph in PDF format to a folder with insufficient rights for file creation, Canoco application crashed. When exporting in EMF format under same circumstances, the failure to store the file went unnoticed.