CAIC is a freeware comparative analysis package for Macintosh. It and finds and calculates phylogenetically independent contrasts in one or more variables enabling you to test hypotheses of correlated evolution.

CAIC was written by Andy Purvis in 1991 using Borland Turbo Pascal at the Evolutionary Biology Group at Oxford University. CAIC 2.0 was developed with Andrew Rambaut using Think Pascal 4.5 (Symantec). Version 2.6.x, has been produced by Nick Isaac and Paul-Michael Agapow using CodeWarrior Pro 4 (Metrowerks) here at Imperial College. If you publish analyses that use CAIC, please cite the following paper :

A. Purvis and A. Rambaut (1995), "Comparative analysis by independent contrasts (CAIC): an Apple Macintosh application for analysing comparative data." CABIOS 11 (3): 247-251.

CAIC is distributed as a binhexed, self-extracting archive and includes two helper programs for creating phylogenies. Alternatively, programs such as TreeEdit may be used to translate from other formats, such as NEXUS. Note certain issues of TreeEdit-CAIC compatability have yet to be resolved.

Two algorithms may be implemented, allowing either nested and non-nested analysis. Text output includes full details of the contrasts and nodal values, enabling the user to check for violations of the model assumptions. Newer versions of the program (version 2.6.0 and later) performs regression through the origin and assumption checks on the output contrasts when two columns of data are selected.

Known bugs

I am grateful to anyone reporting genuine bugs in CAIC. Please include as much information about the circumstances of the bug and, where possible, the program's behaviour in other versions.

CAIC 2.6.5 was the first version capable of conducting statistics on very large datasets. Previous versions contained a memory bug which manifested itself by mysteriously crashing after calculation of around 300 contrasts when two variables were selected for analysis. CAIC 2.0 did not suffer from this problem since no it contained no statistics.

CAIC 2.6.6 contains a bug fix affecting the calculation of contrasts in the Brunch algorithm when the main predictor is continuous. In earlier versions (including v2.0.2) the contrasts were unstandardised when two variables were selected for analysis (wrong) but standardised when there were more than two (correct). This means that output statistics using continuous characters in Brunch were unreliable. This has been rectified so that Brunch always standardises all the contrasts except those of the main predictor variable when that variable is categorical.

CAIC 2.6.7 contains two important bug fixes from v2.6.6. The most serious, spotted by Tim Blackburn, allowed extra contrasts to be generated when operating in the statistics mode (i.e. 2 columns of data). The second bug occurred when multiple columns of data were selected and the main predictor was continuous. The output appeared normal but only contrasts at polytomies were standardised. Neither bug had any affect in Crunch and neither affected any version other than CAIC 2.6.6.

CAICs 2.6.5 through 2.6.7 contained a minor bug which sometimes lead to the value of the largest negative being incorrectly reported. This was fixed in v2.6.8.

CAIC 2.6.8. contains a bug in the calculation of contrasts at large polytomies. This was fixed in v2.6.9.

Technical Details

CAIC runs on any Macintosh running System (> 7.5), with very modest disk and memory requirements (4 MB RAM). Behaviour on other versions of MacOS have not been explored. Note that CAIC cannot run off any read-only media like a floppy disk, CD-ROM or a file server without write access.

CAIC won't run on Intel Macs. They should run in Sheepshaver, but CAIC, MacroCAIC and all other the evolve software causes it to crash.

The existing download package is a self-extracting archive (a Classic application). This means the download basically doesn't work on a PC or Intel Mac and looks like a fault with the software. The zip and distribution files can be opened universally, but may not work.

The dataset can contain up to 10,000 taxa and 128 columns of data. Up to 20 columns of data may be selected in any one analysis. Each tip must be separated from the root of the phylogeny by no more than 49 nodes.
Due to improvements in the software, differences may exist between the latest version of the software and that described in the manual. Many are described in the version history, others are merely cosmetic.

CardInput is a application for building the phylogenies used by the CAIC suite of programs. This latest version fixes some limitations and improves error detection. It is now included in the download packages of CAIC and MacroCAIC.