Download R version 2.9.1 from our UK site or our US site, this is a free package. Once downloaded, install and use all default settings.
We have tested all techniques in Brodgar with R version 2.9.1. Using a newer R version, you may (or may not) encounter problems. Multiple R versions can be installed on your computer; they do not interfere with one another.
Download Brodgar from: www.brodgar.com/setup272.exe
If you did not install R, then please do this now. Install R on your computer by clicking on the R installation file that you downloaded.
Install Brodgar by clicking on the Brodgar setup file. Install under C:/Brodgar and NOT under C:/Program Files/Brodgar.
Start Brodgar by clicking on the Brodgar shortcut on your desktop. The first time you run it, a window will appear asking you for the two license codes we emailed.
Follow steps 1 to 3, when Brodgar asks for the two license codes, just click on the OK button, Brodgar will go into evaluation mode and allow you to work with a large set of internal data sets which were also used in Analysing Ecological Data by Zuur et al. (2007). The evaluation mode is a working version, but importing data is not allowed.
Join our (moderated) mailing list for all updates, new features, technical Brodgar questions and information on courses, etc.: http://lists.highstat.com/mailman/listinfo/brodgar
The Brodgar manual is available online.
|
Manual |
||
| pdf document | Remarks | |
| Cover | ||
| Table of contents | TOC | |
| Chapter 1: Introduction | Chapter 1 | Updated: 28-January-2011 |
| Chapter 2 | Chapter 2 | Updated: 1 August 2011 |
| Chapter 3 | Chapter 3 | Updated: 24-July 2009 |
| Chapter 4 | Updated: 11 September 2009 | |
| Chapter 5 | Chapter 5 Power analysis supplement | Updated: 6-September 2009 |
| Chapter 6 | Chapter 6 | Updated: 9-September 2009 |
| Chapter 7 | Chapter 7 | Updated: 9-September 2009 |
Brodgar can import approximately half a million data cells. The upper limit for the number of variables in ordination techniques is 2000. The maximum number of time series for dynamic factor analysis is approximately 25 time series.
August 2011. Release of version 2.7.1
The online verification system has been replaced by a more user-friendly system
Small bug fixes and improvements have been added.
March 2010. Release of version 2.6.6
Fixed a bug related to GLM and using the offset.
Online license verification has been implemented.
September 2009. Release of version 2.6.4
Fixed the last bugs in Brodgar due to upgrading R version 2.9.1. Repeated LOESS smoothing is working again.
September 2009. Release of version 2.6.3
Fixed a few bugs in the data exploration.
To work on diversity indices, you now need to go to: Data Exploration - General Tools - Select a diversity index and click on the Go button. Then click on Index Functions from the menu, copy the diversity index to the clipboard, and add the index to the spreadsheet via: Data Exploration - Change Data - Column - Add variable from clipboard - Continue - Save Changes and Finish Data Import Process.
September 2009. Release of version 2.6.2
You can now copy and paste residuals (or other info) in the Brodgar spreadsheet. This allows you to apply all the data exploration tools on residuals. Missing values (if present in the original data) are added to the residuals.
Brodgar is now linking to R version 2.9.1. This means better GAMs and a better negative binomial distribution in the GAM. The use of the by command in smoothers has been modified according to R version 2.9.1
All help files have been updated.
The license system has been modified so that it now needs your hard disk.
September 2009. Release of version 2.6.1
With one mouse-click you can now visualise the fitted values of a linear regression o GLM model. For example, linear regression models that contain interactions have fitted lines that cross.
July 2009. Release of version 2.6.0
Added more demonstration data sets. These data sets were also used in Zuur et al. (2007; 2009).
July 2009. Release of version 2.5.9 (2.5.8 was used during classes)
Added the multi-panel dotplots (see Dotplot - Settings).
Changed default settings and menu options in the pairplot.
In a scaterplot, you can now use labels instead of points (see the Settings).
Minor improvements.
November 2008. Release of version 2.5.7
Added the negative binomial distribution to GLMs and GAMs.
GLS, mixed effects and GAMM routines have been further improved. See the online help files how to use them.
Updated the manual.
Updated some help files.
December 2007. Release of version 2.5.6
Small bug fix for multivariate regression trees.
November 2007. Release of version 2.5.5
Small improvements and bug fixes to the GAM GUI.
November 2007. Release of version 2.5.4
Fixed an annoying little bug in the GUI for regression, GLM and GAM.
October 2007. Release of version 2.5.3
Took a while (because of the release of our book).
The GAM section has been updated. You can now use the 'by' command. This is useful for interaction between smoothers and nominal variables. We also implemented shrinkage smoothers. These can have 0 degrees of freedom. Useful if many smoothers are not significant; drop them all at once if the estimated degrees of freedom is 0.
Because the mgcv library is excellent for GAMs, we decided to drop support for the gam library from Hastie and Tibshirani. Give a shout if you want it back.
Brodgar should now work with later R versions. Give a shout if it doesn't.
VIF values have been added (useful for detecting collinearity).
Now we will update the manual, and provide text on the GLS, mixed modelling and GAMM functions.
October 2006. Release of version 2.5.2
Model validation tools for dynamic factor analysis have been improved.
SCoTLASS has been added to PCA.
Full model selection (not just forwards and backwards) has been added to linear regression and GLM.
Various small bug fixes in data exploration.
Chapters 1-3 of the manual have been corrected on grammar.
Other chapters are nearly finished.
March 2006. Release of version 2.5.1
VIF values have been added to linear regression and the data exploration. VIF values can be used to detect collinearity.
December 2005. Release of version 2.5.0
Generalised least squares (GLS) has been added. GLS is like linear regression except that an auto-correlation structure can be used for the residuals.
The GUI for mixed modelling has been simplified and the help files have been updated.
Fixed a little bug for chronological clustering if very large values are used (thanks to Pablo Almaraz García).
Fixed a bug in the BVSTEP procedure.
December 2005. Release of version 2.4.9
The bug that occasionally didn't allow R graphs to maximize has been fixed. The user can only open one window at a time.
PCA Equilibrium circles have been added.
Order of lattice plots can be changed
Power analysis has been added.
General small improvements.
Thanks to Dr. Borges: fixed a little bug in regime shift analysis.
October 2005. Release of version 2.4.8.
60 MSc students in Aberdeen have worked with Brodgar for 3 weeks and managed to find a few little bugs. Thanks! There is still this little annoying bug in Brodgar that occasionally doesn't allow R graphs to maximize.
A new spreadsheet editor and viewer have been added.
Numerical output is now routed via our own data editor.
Violin plots have been added.
The Brodgar mailing list is now operational.
The price structure has changed as more and more universities are buying Brodgar and request licenses for student home use.
2.4.8: Fixed a little bug in the MDS and ANOSIM code in version 2.4.7.
July 2005. Release of version 2.4.5 and 2.4.6 (August).
Fixed a little bug in the tree GUI.
July 2005. Release of version 2.4.4.
Various improvements to discriminant analysis have been added. Groups of samples can be identified using a (explanatory) variable.
A bug fix for the GAM and GLM graphical user interface.
A spreadsheet interface to simplify the data import process is in preparation.
ANOVA and power analysis is in preparation.
Preparations for nested GLM and GAM models have started.
Preparations for an attached text editor have started.
May 2005. Release of version 2.4.3.
Some catching up with PRIMER: ANOSIM, BIOENV and BVSTEP modules have been added.
The Mantel test and partial Mantel test have been added.
Mixed modelling has been implemented.
March 2005. Release of version 2.4.2.
Small updates.
February 2005. Release of version 2.4.1.
Dotplots can now be made conditional on nominal variables.
Extra pairplot styles have been added.
January 2005. Release of version 2.4.0.
The recent modifications caused a little bug in the interaction plot; fixed.
The data import process has been improved and is now considerably faster. Variable names that might give errors for the R tools (e.g. Y%1 or Y-1) are automatically adjusted.
December 2004. Release of version 2.3.9.
More help text for AIC values of GAM models have been added. This should clarify how the AIC values are calculated.
Font sizes of graph labels and titles of R graphs can be changed. See the Tools menu.
December 2004. Release of version 2.3.8.
Automatic detection of R installation directory has been implemented.
Bug fix for quasiPoisson and quasiBinomial GLM and GAM models.
AIC values obtained by GLM and GAM using two slightly different formulae are presented.
November 2004. Release of version 2.3.7.
General improvements.
November 2004. Release of version 2.3.6.
A link to the R library gam has been added. A so-called "Specialised corner" was added to gam allowing for easy implementation of forward and backwards selection procedures to find the most optimal model, and models with different degrees of freedom.
Small bug fixes and general improvements.
Brodgar now needs R version 2.0!
October 2004. Release of version 2.3.5.
A link to the R library mvpart for multivariate tree models has been added.
Fixed a little bug in the canonical correlation analysis function. Thanks to Dr. Dobrev.
Non-metric MDS has been added.
Most measures of similarity discussed in Chapter 7 of Legende and Legendre (1998) have been added, see the MDS function.
August 2004. Release of version 2.3.4.
GAM, GLM and tree models now allow for weights.
Fixed a little bug in the AIC for additive modelling.
July 2004. Release of version 2.3.3.
Added further time series techniques (lattice graph of auto-correlation functions and lattice graph of different span widths in repeated Loess smoothing).
June 2004. Release of version 2.3.2.
Minor bug fix for the scaling options in RDA and CCA.
June 2004. Release of version 2.3.1.
Minor bug fix for plotting the fit in GLM and regression models if only one explanatory variable is selected.
June 2004. Release of version 2.3.0.
Interactions between variables in linear regression, GLM and GAM have been added.
Three time series techniques have been added.
Significance testing for CCA and partial CCA are now included.
Significance testing for db-RDA transformations has been added.
Design and interaction plots have been added. Brodgar now needs R version 1.8.1 (which can be downloaded from our site).
April 2004. Release of version 2.2.8.
ARIMAX modelling has been improved.
March 2004. Release of version 2.2.7.
Principal coordinate analysis (PCO) was added to chronological clustering.
Added an extra example on chronological clustering and PCO to the website (see the time series section).
March 2004. Release of version 2.2.6.
Chronological clustering was added.
Scatterplots can now have multiple lines (conditional on nominal variables).
February 2004. Release of version 2.2.5.
Pair plots can now be made for Y and/or X.
Fixed a nasty bug in the (partial) RDA/CCA triplots. If explanatory variables were deselected in the multivariate analysis menu, names of nominal and non-nominal variables were mixed up.
February 2004. Release of version 2.2.4.
Fixed a little mistake in the export facilities of canonical correlations of MAFA.
January 2004. Release of version 2.2.3.
Fixed a little bug in the similarity ratio (Measures of similarity).
Updated the first 6 chapters of the manual.
December 2003. Release of version 2.2.2.
Options to use "semi-parametric models" have been added to additive modelling (AM) and generalised additive modelling (GAM).
More graphical validation techniques have been added to AM and GAM.
Chapter 5 of the manual has been considerably modified.
December 2003. Release of version 2.2.1.
More graphical validation techniques have been implemented for linear regression and GLM.
There was a little bug in the species richness index function (Exploration - General tools - Diversity index - Species richness).
R graphs can now be exported to EPS (which is useful if you write a book).
All R graphs under "Exploration" and "Univariate" are now presented in the same way. The "Multivariate" and "Time series" tools will be modified later this month to follow the same format.
December 2003. Release of version 2.2.0.
To view, print and export R graphs, a third party graphical software package (Irfanview) has been added to Brodgar.
Linking R and Brodgar has been simplified.
Smoothing and regression options have been added to the lattice and pairplots.
More index functions have been added to the data exploration functions.
Cosmetic changes have been made to chapters 1 and 4 of the manual.
November 2003. Release of version 2.1.9.
Implementation of Bray-Curtis ordination (also called: Polar ordination).
Histograms: number of bars can be selected.
Alternative db-RDA transformations have been added to PCA and CA.
More options for plotting the regression and classification trees.
Two new chapters of the manual are now online.
Work in progress: (I) Tests to add a third-party graphical package to open and edit "R-generated-graphs" have started. (II) Linking db-RDA transformations with significance testing. (III) Extending the significance testing facilities from RDA to CCA.
November 2003. Release of version 2.1.8.
Implementation of classification trees.
Tree pruning in regression and classification trees can now be done using the cp parameter.
Chapter 5 of the new manual has been uploaded. It contains various examples on linear regression, partial linear regression, additive modelling and regression/classification trees.
October 2003. Release of version 2.1.7.
Implementation of conditional histograms (see Exploration -R tools).
October 2003. Release of version 2.1.6.
Bug fix in scatter plots and coenoclines.
October 2003. Release of version 2.1.5.
Fixed a bug in the percentage similarity function. Thanks to Pam Sikkink.
October 2003. Release of version 2.1.4.
Version 2.1.3: The graphical user interface for discriminant analysis had a little (but fatal) bug. Fixed it.
Forward selection of explanatory variables in redundancy analysis (RDA) and canonical correspondence analysis (CCA) has been implemented. permutation tests for (partial) RDA have also been implemented. This will give a p-value for the conditional effects of explanatory variables (or axes). Results were compared with CANOCO version 4, and are identical (except for differences due to a different random seed).
Using "Legendre and Gallagher (2001). Ecologically meaningful transformations for ordination of species data, Oecologia, 129:271-280", an alternative to distance base RDA (dr-RDA) has been implemented.
October 2003. Release of version 2.1.3.
A "Getting started" menu button has been added.
The structure of Brodgar has been changed to: Import Data - Exploration - Univariate analysis - Multivariate analysis - Times Series analysis.
The user can now select response and explanatory variables from within the multivariate techniques.
The manual has been partly revised (and corrected on grammar).
Programming on permutation tests for RDA and CCA has started. The same holds fro db-RDA.
July 2003. Release of version 2.1.2.
We used Brodgar in our statistics course in Faro, Portugal. Based on feedback from 21 course attendants, various modifications to the R interface were made. Further modifications are in progress.
A English grammar correction of the manual is in progress.
July 2003. Release of version 2.1.0.
The following time series methods have been added: time series regime analysis, ARIMAX, spectral analysis.
Various improvements to the R interface.
May 2003. Release of version 2.0.9.
GAM and GLM have been extended with binomial and Gamma distributions. Quasi likelihood methods are also implemented.
Added credit card payment facilities.
Fixed one little bug.
Updated chapters 1 to 4 of the manual!
April 2003. Release of version 2.0.8.
Boxplots, coplots and lattice graphs can be made conditional on two nominal variables.
Graphs produced by the R interface can be saved as emf files. These are high quality graphs which can be imported into Word.
A few small bug fixes.
April 2003. Release of version 2.0.7.
Partial regression, regression and scatterplots have been added to the R interface.
The Brodgar website and Chapter 9 of the manual have been extended with a partial linear regression example and a clustering example.
Fixed a bug in the regression and GLM functions (the automatic backward selection method failed if the data contained missing values).
April 2003. Release of version 2.0.6.
Interface to R has been added.
Uploaded: Chapter 9 of the manual (Brodgar and R).
February 2003. Release of version 2.0.5.
Bug fix in GUI for partial CCA and partial RDA.
First trials to link R and Brodgar are successful.
December 2002. Release of version 2.0.4.
Partial RDA has been implemented.
Variance partitioning example has been added to our web site.
Brodgar results obtained by CCA, RDA, partial CCA and partial RDA were compared with CANOCO 4.0. Results are identical.
Zooming facilities of graphs has been added. Use the left mouse button to zoom in, and the right mouse button to zoom out.
November 2002. Release of version 2.0.3.
Partial CCA has been implemented. This means that variance partitioning can be applied (estimate the pure spatial effect, or the pure management effect). Partial RDA will come soon.
Diagnostic tools (e.g. quality of display along the first two axes per response variable) have been implemented.
More flexibility for biplots and triplots. The user can now choose how to present response variables, explanatory variables and samples in biplots or triplots; e.g. by labels and lines, labels, squares, circles or triangles. It is also optional to add axes through the origin.
October 2002. Release of version 2.0.0.
The data import process has been modified and considerably improved. It is not possible anymore to import data with empty cells. This will avoid a lot of vague error messages.
Part of the import data calculations are now carried out in FORTRAN. For large data sets, this should reduce the calculation time by 30-40%
Imported data are now internally stored with only two decimal precision (instead of 8).
Additional transformations and standardisations have been implemented.
Chapters 1, 3, 4 and 5 of the manual were revised.
Selection of response variables and explanatory variables has been simplified (try Shift-Mouse move).
August 2002. Release of version 1.9.2.
A user in Finland encountered some problems with a semi-large data set (70 response variables and 1000 samples) and canonical correspondence analysis. Problems were related to a matrix multiplication (the dimensions became rather large). We have improved the code and the problem is fixed.
August 2002. Release of version 1.9.1.
Implementation of canonical correlation analysis and Webster's method to detect discontinuities.
July 2002. Release of version 1.9.0.
Implementation of MAFA.
Various improvements to the discriminant analysis GUI.
Fixed a few bugs in the GUI.
June 2002. Release of version 1.8.9.
Small modification in presenting high correlations.
June 2002. Release of version 1.8.8.
New graphical options in "Plot data".
Small modifications in the dynamic factor analysis techniques.
June 2002. Release of version 1.8.7
Bug fix and modification of the code for coenoclines. The code is now approximately 10 times as fast. Titles and the range of the y-axis in coenoclines along explanatory variables can now be changed.
May 2002. Release of version 1.8.6.
Bug fix in the cross-correlation function.
Titles, labels and fonts for time series plots are still not working properly. Sorry. Will be fixed in the first week of June.
January 2002. Release of version 1.8.5.
Two more data examples have been added, namely
the hunting spider data. This data set has been used in many textbooks to illustrate the use of CA and CCA to analyse species-environmental relationships.
Fisher's Iris data. A data set frequently used to demonstrate discriminant analysis.
Projects (and associated directories and files) can be deleted.
Biplot and triplot titles, labels and fonts can be stored and retrieved.
Besides ordinary standardisation of explanatory variables in CCA, weighted standardisation (default settings in CANOCO) can be chosen. See the help file for further details.
June-December 2001. Release of version 1.8.1 to 1.8.4
Mainly bug fixes for Windows NT and 2000 operating systems.
Titles, labels and fonts for the triplots and biplots can be changed.
May 2001. Release of version 1.8
Help files have been added.
Brodgar has been extended with discriminant analysis and factor analysis.
The manual is being revised.
The data import process has been simplified.
March 2001. Release of version 1.7
Minor bug fixes
In dynamic factor analysis, the covariance matrix of the error component can now be modelled as (i) a diagonal matrix or (ii) a symmetric, non diagonal matrix.
Chapters 5, 6 and 7 of the manual are available online.
February 2001. Release of version 1.6
Generalised Procrustes analysis (GPA) has been added. Suppose the data consist of M samples of N variables made on T occasions. This is called 3-way data. Two examples are (i) N species measured at M sites in T years and (ii) N panel members assessing the quality of M products, during T assessments. GPA can be used to analyse such data. It applies a dimension reduction technique on each 2-way data set. In the first example, there are T tables, each containing species-by-sites data. GPA uses the T ordination diagrams to calculate an average ordination diagram. An analysis of variance indicates how well (i) individual species, (ii) each of the ordinations and (iii) each of the axes are fitted by this average ordination diagram. GPA can be used to analyse very short time series.
A first set of help files have been added.
January 2001. Release of version 1.3
The layout of the GUI has been improved considerably.
Implementation of the following dimension reduction techniques:
Redundancy analysis. This is basically a principal component analysis in which the axes are restricted to be linear combinations of the explanatory variables.
Correspondence analysis.
Canonical correspondence analysis (this is a correspondence analysis in which the axes are restricted to be linear combinations of the explanatory variables).
Auto- & cross-correlation functions have been improved.
Various dynamic factor models have been added.
Limitations on dimensions (number of time series and length of time series) of the data have been removed. The maximum number of common trends has been set to 10. Please note that the algorithm for dynamic factor analysis becomes unstable if more than 30 time series are used.
November 2000. Release of version 1.2.
November 2000. Release of version 1.1.
November 2000. Version 1.0 is released.