Canoco 5 Frequently Asked Questions

Questions

  1. Is there an upgrade available for users of older versions?
  2. Does Canoco 5 run on Windows 8/10 and other Windows platforms?
  3. Does Canoco 5 run on Linux or MacOS?
  4. Is Canoco 5 a 64-bit program?
  5. How does Canoco deal with the stability issues reported by Oksanen & Minchin (1997)?
  6. Does Canoco 5 provide NMDS?
  7. Is Canoco 5 able to analyse my data set with 20,000 samples and 600 species?
  8. How does Canoco 5 compare to PC-ORD version 6?
  9. How does Canoco 5 compare to the Vegan package in R?
  10. I bought Canoco 4.x at Scientia Software, which is no longer distributing Canoco. Whom should I ask for technical support or upgrade?
  11. How do I omit cases or response variables from an analysis?
  12. How can I display ordination score values?
  13. Why is the transformation of response data disabled in the Analysis Setup Wizard?

Answers

1. Is there an upgrade available for users of older versions?

Yes, you can upgrade from older versions of Canoco. Please, contact the distributor from whom you bought your original Canoco version to get more precise information on availability and pricing (Scientia customers, please see the question 10).

2. Does Canoco 5 run on Windows 8/10 and other Windows platforms?

Canoco works fine on standard desktop and laptop computers with Windows 8, 8.1 or 10 for both the 32-bit and 64-bit versions, available at the time Canoco 5 was released. Window 8.x and 10 also support other platforms, such as small mobile devices managed by touching the screen with your fingers, but, as any program analyzing and visualizing scientific data, Canoco 5 will not excel on them.

Canoco 5 works on other Microsoft operating systems as well, starting from Microsoft Windows XP with SP (service pack) 2 or SP3 installed. This includes also Windows Vista and Windows 7.

3. Does Canoco 5 run on Linux or MacOS?

There is no native implementation of Canoco 5 for those operating systems, but Canoco 5 was tested to run under the Wine environment on Linux and likewise in a similar environment of CrossOver package. The following snapshot illustrates working with a biplot diagram in Ubuntu Linux workspace.

 

4. Is Canoco 5 a 64-bit program?

There is now just one application within Canoco 5 package and the installer installs its 64-bit version on 64-bit Windows and the 32-bit version on 32-bit Windows. The files read and stored are compatible between 32-bit and 64-bit versions of Canoco 5.

5. How does Canoco deal with the stability issues reported by Oksanen & Minchin (1997)?

The bug in the original DECORANA (Hill, 1979) code for Detrended Correspondence Analysis has been studied and was repaired before Canoco 4 version, 15 years ago. It turned out that the change suggested by Oksanen and Minchin (1997) did not make the smoothing routine order-invariant. Another change was required as well, as agreed upon by Oksanen & Minchin in subsequent correspondence (see  http://cc.oulu.fi/~jarioksa/pages/isbug.htm).

6. Does Canoco 5 provide NMDS?

Yes, the non-metric multidimensional scaling (NMDS or NMS) method is now fully integrated into Canoco 5, replacing the add-on program WinKyst from the previous version. The WinKyst program remains available as a free alternative solution for performing NMDS analyses.

In addition, Canoco 5 provides a wide array of ordination methods that operate directly on the data, instead of on derived similarity or dissimilarity (distance) measures, as NMDS does. The advantage of these methods over NMDS are that:

  • The resulting ordination readily provides information on the level of individual variables in the data table
  • The ordination can be focused on the effect of particular explanatory variable(s) – in the form of constrained ordination – so that it shows e.g. the community response to explanatory variables
  • Background variation can be removed from the ordination by using covariate data, so that the ordination can display new information rather than trivial or uninteresting variation, or variation that is already well understood.

7. Is Canoco 5 able to analyse my data set with 20,000 samples and 600 species?

Yes, most certainly it is, and much larger data as well! There are always limits to the size of data one can analyze on a personal computer, but in version 5, Canoco itself enforces only the limit of 16.7 millions table rows and 16.7 millions table column, which is not very limiting. The operating systems, however, impose much harsher limits, namely in the case of 32-bit Windows. Standard applications cannot allocate in 32-bit Windows more than 2 GB of memory in total, but Canoco 5 uses a special provision to extend its data space to 3 GB (after proper Windows re-configuration). A single, contiguous memory block cannot be allocated with a size greater than ca 1 GB, so Canoco 5 data tables are limited to roughly 134 million entries. Yet another limitation usually comes from the processor speed, making analyses of really big data unbearably slow (but you can leave the computer working overnight, then).

8. How does Canoco 5 compare to PC-ORD version 6?

Canoco 5 is software primarily focused on ordination methods, while PC-ORD offers a wider range of methods, including cluster analysis, TWINSPAN, species-area curves, Mantel test and similar distance-based permutation tests.

We believe, however, that due to its focus on ordination, Canoco does a better job in this field both in terms of offered statistical methods and user friendliness of the software and produced graphical output. PC-ORD version 6 (as of October 2012) does not support partial constrained and unconstrained analyses, needed to address many of the non-trivial research questions of today. Consequently, PC-ORD has neither support for selecting a subset of explanatory variables reasonably describing the community variation, nor the ability to easily evaluate simple as well as conditional effects of individual predictors collected in the project. Additionally, PC-ORD does not support correct permutation tests for datasets coming from several frequently encountered design types, including repeatedly measured permanent plots or data coming from linear transects or rectangular sampling grids.

Naturally, both programs continue to develop. Version 6 of PC-ORD added linear constrained ordination (redundancy analysis - RDA) that was missing in previous versions, while Canoco 5 adds the non-metric multidimensional scaling (NMDS) that was previously available only as a (free) add-on package. Canoco 5 features the symmetric  form of co-correspondence analysis to relate two biotic communities together. Both Canoco 5 and PC-ORD are valuable analytical tools which can be used to corroborate one another where they offer overlapping functionality. Each complements the other with analyses not available in the other package.

We would like to mention here some of the gross mistakes in the PC-ORD – Canoco comparison table published on PC-ORD web site (http://home.centurytel.net/~mjm/canoco.htm as of October 2012). This comparison concerns the previous version of Canoco (4.5), so newly added features of Canoco 5 are naturally not covered (no strict size limits either in analyses or in graphing, import from CSV files, presence of NMDS, storage of distance matrix), but some of the claims were not correct even for version 4.5 and may be perpetuated into any new comparisons.

Most important is the comparison of prices, where MjM company boasts about a substantially cheaper product, but the comparison for a single user license compares a personal license of PC-ORD with a Canoco price for private companies. A fairer comparison would be among educational licenses purchased by an academic department for its employee. As the PC-ORD pages tell you, if the license should stay at the institution which bought it for you, the price is – for a single user – 299 + 100 USD, which does not compare so well with a single educational  license of 368 USD for Canoco 5 (same as for Canoco 4.5). Similarly, the student licenses are offered for Canoco as additional educational licenses for 238 USD, which does not compare so badly with the 199 USD for PC-ORD. It is only fair to add that shipping costs are smaller for PC-ORD.

As a rather minor point, we shall note that Canoco 4.5 had (and continues to have in version 5) the "successional vectors" called series collections in PC-ORD, that monotonic transformations (namely log and sqrt) are available, as are the diversity indices computed for individual cases, and the detection of outliers with leverage statistics for individual predictors as well as for their whole set.

9. How does Canoco 5 compare to Vegan package in R?

The vegan package in the R program implements most of the methods available in Canoco 5 as well as a few more. But as R is an open system, those Canoco methods missing in vegan can be found in other packages, related or unrelated to vegan. However, the R environment is primarily command-oriented and to import data, perform analysis or create simple ordination diagrams, commands calling functions are needed, often with a complex set of function parameters to correctly choose from. We see the R environment as excellent for a statistically-oriented ecologist, but for researchers focused on their research subjects, using vegan in the R environment has a steep learning curve, which must be climbed repeatedly for someone evaluating their data after longer periods of work in field or lab. There is a relatively user-friendly interface in the Rcmdr package, offering in combination with the DiversityR package some of the multivariate methods of vegan, but only for a small subset, excluding all the advanced methods and advanced options.

10. I bought Canoco 4.x at Scientia Software, which is no longer distributing Canoco. Whom should I ask for technical support or upgrade?

Please ask your support or upgrade questions at this address:  This email address is being protected from spambots. You need JavaScript enabled to view it.

 

11. How do I omit cases or response variables from an analysis?

When you start to use Canoco 5 on a particular user account, the Analysis Setup Wizard is shown in so-called QuickWizard mode. This is reflected in the QuickWizard button (emphasized by a red rectangle in the following snapshot) being shown in depressed state.

Whenever you want to omit (or weight) cases or response variables, you must switch the QuickWizard mode off by clicking once this button (its look then becomes flat), before creating the new analysis (e.g. clicking the New button below the list of analyses) or before changing options for an existing analysis (e.g. by clicking the Modify button and then selecting Replace in existing analysis).

With QuickWizard mode switched off, the Analysis Setup Wizard shows more pages, including the Response and Explanatory Data page:

If you check (as in the above snapshot) one or both boxes labelled "Select"..., the setup wizard will later show corresponding pages where a subset of cases and/or response variables can be selected for the analysis.

To switch the QuickWizard mode on again, you click the same button in the toolbar (it is a toggle button).

You can define groups of cases and of variables using the Project | Groups | of ... menu commands and then select or omit cases or variables belonging to a particular group by using the From group button on the respective item selection page of the setup wizard.

12. How can I display ordination score values?

The ordination scores and additional ordination statistics that could be obtained in CANOCO 4.x versions from a solution (.SOL) file are still available in Canoco 5 analyses, but normally they are not shown. When you start using Canoco 5 on a particular user account, the analysis notebooks are shown in a brief mode with only the summarizing pages and created graphs available, the other information (ordination scores and analysis log) being hidden. To display analysis notebook in a non-brief mode, you must select the Edit | Settings | Canoco5 options command from the main menu and in the General page of the Canoco Options dialog box uncheck the "Show brief version of notebooks with analysis results" box:

 - closing then the dialog with OK button.

Please note that already displayed notebooks do not change its contents after this change in options. You must close the notebook (e.g. using the Hide button below the list of analyses) and then reopen it again (e.g. using the Show button that replaces the Hide button for an analysis with closed notebook). When hiding and re-displaying analysis notebook, you should note that if there are multiple analyses defined and an analysis notebook is hidden, the focus often switches to another analysis. So to re-open the analysis notebook for the original analysis, you must re-select it in the list of analyses.

The scores shown in the score pages can be copied to Clipboard and further reused within or outside of the Canoco 5 program. There are two context menu commands for copying selected score columns:Copy scores copies the table contents as seen in the score page, whileCopy expanded aligns the set of rows with the corresponding set of data items (cases, response variables, etc.) present in the project: this is important when you want to copy computed scores into the project as a new data table (some data items might be absent in the analysis output, factor variables on the other hand generate more than one set of scores).

 

13. Why is the transformation of response data disabled in the Analysis Setup Wizard?

You might need to execute principal components analysis (PCA) on your data, for example, and you want to log-transform the values of the variables used in this PCA. But when you arrive to the appropriate setup wizard page, the log transformation is not available:

 

Why is it so? There are essentially two possibilites:

(a) Your data table is incorrectly specified as a general type, while it is in fact compositional one. Data table type was specified at the time of data import and you have perhaps made a mistake. The most important property of a compositional data table is that all its variables (columns) are measured in the same units and it makes a good sense to sum them up within each case (table row). For example, each column can be a biotic species and the values are abundances or biomass estimates for each species within each case. Or the columns can be different types of phospholipid fatty acids (PLFA) and the values are their concentrations in different samples. If this is your case, you can change the table type by selecting the table in the upper left area of Canoco 5 (as Table 2 in the following snapshot) and then choosing the Data | Change table kind to ‘compositional’ menu command.

 

(b) Your data table is correctly specified as a general type. Different columns of such data table are measured in different units or – if the units are identical for all of them – it still does not make sense to sum them up. This is the case with many data sets characterising the environment in which a community was sampled (e.g. water analysis results, soil chemistry analysis results etc.). In such a case, you can still specify transformations of the variables, but you must do it for each of them separately (because, likely, the chosen transformations will differ across the variables). These transformations can be specified by selecting the concerned data table in the upper left area of the Canoco workspace and then the Data | Default transformation and standardization menu command from the Canoco 5 menu. This command displays following dialog box where you can select individual variables (except the factor variables) and then choose an appropriate transformation. In fact, Canoco Adviser tries to suggest transformation type, but its scope is limited to the choice between “No transformation” and log-transformation (suggesting appropriate constants in the latter case).

 

With the default transformations set, they are used always when the concerned variable is used in an analysis, whether you use it as a response variable or as a predictor variable (explanatory, supplementary or covariate). To make sure this happens, you can either check back at the Variable Transformations dialog box or you can inspect the Log tab of the analysis notebook (see the previous FAQ 12, How can I display ordination score values, for the description of how to display analysis notebook in a non-brief mode).