.\"
.\" abxtest - simple ABX double-blind testing script
.\" Copyright (C) 2000-2004 Robert Leslie
.\"
.\" This program is free software; you can redistribute it and/or modify
.\" it under the terms of the GNU General Public License as published by
.\" the Free Software Foundation; either version 2 of the License, or
.\" (at your option) any later version.
.\"
.\" This program is distributed in the hope that it will be useful,
.\" but WITHOUT ANY WARRANTY; without even the implied warranty of
.\" MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
.\" GNU General Public License for more details.
.\"
.\" You should have received a copy of the GNU General Public License
.\" along with this program; if not, write to the Free Software
.\" Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
.\"
.\" $Id$
.\"
.TH ABXTEST 1 "16 November 2000" "MAD" "MPEG Audio Decoder"
.SH NAME
abxtest \- double-blind ABX comparison testing script
.SH SYNOPSIS
.B abxtest
.RB [ -n
.IR min ]
.RB [ -m
.IR max ]
.RB [ -g
.IR goal ]
.I A-cmd B-cmd
.SH DESCRIPTION
.B abxtest
is a tool for conducting listening (or other subjective) tests to determine
whether a listener can discern a difference between two subjects under test,
denoted
.B A
and
.BR B .
.PP
The listener conducts several trials in which an unknown subject,
.BR X ,
is randomly chosen to be either
.B A
or
.BR B .
The probability that
.B X
is
.B A
or
.B B
is the same. The listener must decide whether
.B X
sounds the same as
.B A
or the same as
.BR B .
The listener may repeat
.BR A ,
.BR B ,
and
.B X
as many times as necessary before reaching a verdict.
.PP
In order to prove the hypothesis that a difference can be discerned between
.B A
and
.BR B ,
the listener must disprove the null hypothesis that
.B A
sounds the same as
.BR B .
This is accomplished by correctly identifying
.B X
often enough that randomly guessing correctly the same number of times would
be statistically improbable.
.PP
The
.B A
and
.B B
subjects for
.B abxtest
are actually arbitrary shell commands. The normal mode of use would be to give
commands which play two slightly different sounds. The standard output and
standard error from each command are suppressed so as not to reveal the
identity of
.BR X .
The listener may interrupt an executing command (usually by pressing
.BR ^C )
without terminating the test.
.SH OPTIONS
.TP
\fB\-n\fR \fImin\fR
Set the minimum number of trials. Trials will continue after this number if
the statistical probability goal has not yet been met. The default is 10.
.TP
\fB\-m\fR \fImax\fR
Set the maximum number of trials. Trials will end after this number even if
the statistical probability goal has not been met. The default is 20.
.TP
\fB\-g\fR \fIgoal\fR
Set the minimum statistical probability goal. This is the probability that the
number of correct identifications is the same as random guesses. In order to
disprove the null hypothesis, the result of the test must be a probability
less than or equal to this goal. The default is 0.05.
.SH FILES
.TP
.I /dev/random
By default this file is read to generate a random sequence of
.BR X .
.SH ENVIRONMENT
.TP
RANDOM_FILE
If set, the file named by this variable will be read instead of
.I /dev/random
to generate a sequence of
.BR X .
.SH NOTES
.B
abxtest
uses the binomial distribution to accurately calculate the probability that
the result is the same as random guesses.
.SH AUTHOR
Robert Leslie <rob@mars.org>