Denis Cousineau home page

Me

The PDF documents on this site were made with PrimoPDF.
Get a reader free.

The Notebook documents on this site were made with Mathematica.
Get the reader free.

Pages © Denis Cousineau;
last modified: December 17th, 2009;
page consulted on

Programme d'Analyse Statistique de Tendance et de dIStribution

Statistical Curve and Distribution Analysis Software

This program is intended to analyze learning curves, such as the power-law and the exponential curve. It can also return parameters describing distribution of raw data. This program works with parameters naming the input file, its number of column, the column containing dependant variable, and so on. 

For example,

pastis -r data.dat -c 10 -d 8 -a PLAWTREND c2

will analyze the power-law curve of data contained in the file data.dat, which has 10 columns, and where reaction times are in column 8. The session number is contained in column 2.

This software is no longer maintained.

This software is free to download. However, if you use it in a publication, please mention the reference. You may adapt it freely. If you think the enhancement is valuable, keep me informed, I shall adapt the software, and mention your contribution.

Complete reports

the first report describe PASTIS v1.0 and how to use the basic arguments for fitting curves and distributions. The second describe the use of average data on learning curves.

PASTIS: the report on what it is. Format: PDF file; size: 430K, published in: Behavioral Research, Methods, Instruments & Computer, 1997 (nov.) p. 542-548.

Testing curvatures of learning curves on trial and block average data (format PDF, 570K).

Download area: PASTIS v3.02

Binaries are

• available for DOS but not very stable
• available for UNIX, may not work on all platforms

Source code

Source code The source code that can be used to modify the program or to compile PASTIS on other platforms. It is composed of a c main procedure with Fortran library.

Release information

See history.txt for detailed release information.

PASTIS version 2.1: This revision include possibility of constraining a parameter using greater than and lesser than expressions. For example:

will constraint parameter 1 to be between 200 and 400. Beware, on UNIX system, you must use escape sequence before < and >, such as \<.

PASTIS version 2.0: This new version uses another minimization function, STEPIT, written in FORTRAN, and made available by G. D. Logan. Therefore, there is no need to further use a minimization shell, as recommended in the above paper. Furthermore, it optimize the use of ram whenever possible, instead of doing multiple read operation from disk.   Includes the source code written in C and FORTRAN language, and a compiled version.

PASTIS 1.0: The main options are: -r and -w: read from and write to (write to can be replaced by a > pipe -c and -d: number of column and column containing dependant variable -s and -h: (optional) give starting value or hold constant value of a parameter -b and -f: (optional) break input file according to value of a certain column or filter input file. -o, -g and -v: (optional) help, display debug information, use verbose output. -a: start analysis named. Names are: WEIBULL, EX-GAUSSIAN, PLAWTREND, GPLAWTREND, EXPOTREND, and GEXPOTREND. For example,

will analyze data.file, which has 4 column and dependant variable is in the fourth (-c 4 -d 4). Since I put subject # in col. 1, and want separate analysis, I break the analysis according to c1 (-b c1). I hold the value of first and second parameters at 150 and 250 (-h p1=150 p2=250). Finally, I demand a WEIBULL distribution analysis.

Last Modified: January 5th, 2003