Neural Information Processing

Standalone (Command-Line) Executable


The psignifit engine source code can be compiled to run as a command-line executable. On the UNIX or DOS command line, it can receive its input through stdin or through named files, in the manner of most command-line utilities. Under Classic MacOS, the utility takes all its input from a text file called "prefs" in its own directory. On the download page, you can download the standalone utility, ready compiled for various platforms. The archive contains:

Quick-Start Guide

On MacOS 9.x and below, the flexibility of the standalone application is limited. All its input comes from a text document called "prefs" in the application folder. Edit the "prefs" document, then simply double-click on the application to run.

On Windows, the flexibility of the standalone is limited simply because the flexibility of the command-line is limited. It is highly recommded that you download Cygwin, and then use a Cygwin-compiled version of psignifit under the bash or tcsh shells.

As a command-line executable, psignifit can accept input in a number of different ways:

Example dat and prefs are supplied with the standalone distribution. (NB: for the Mac application, the "prefs" file must contain the data too, so there is no separate dat file.

See the enclosed help documents for an explanation of the preferences format, a list of available options and a glossary of common terms. Note in particular the #WRITE_... options described in psych_options.txt), which allow different parts of the output to be redirected to named files, or to the standard output. If no #WRITE_... options are specified, then certain matrices are printed to stdout by default - this is the output you will probably see the first time you run the utility, and the parts you are most likely to be interested in are the TH_LIMS and SL_LIMS matrices, whose format is as follows:

Explanation of the LIMS matrix format

(title denotes that the matrix contains BCa limits on thresholds)
(0.023 confidence level)
(0.159 confidence level) 
(rows denote confidence
(0.841 confidence level)
 levels - default values shown)
(0.977 confidence level) 
(columns denote threshold
levels - default values shown)