This is a beta version (potentially buggy) of ABX software which pays
more than just lip service to the analysis of type II errors.
In the ABX module, critical values for type I and II error risks and
effect size can be entered before starting a listening test. There is
also a training mode so that effect size can be estimated.
Presets are offered for identifying "obvious," "moderate," and
"subtle" differences. These presets are geared towards the comparison
of audio codecs, where false acceptance of "no difference" is not
important (i.e., type II error risk is preset to a rather large 0.2),
and also where a strict p-value is not needed nor even desirable.
That is, the critical type I error risk is preset to 0.05, and not to
something very strict like 0.002 (which is what 14 correct out of 16
would yield for a p-value).
Feedback and bug reports are appreciated.
Known bugs:
1. If ABX window is left open when new WAV files are loaded, the old
status values are not cleared off.
Additional features to be added before release:
1. More information to be written to the results file (i.e., the
critical values for the error risks, the resulting type II error when
a test is finished).
2. If an ABX is successfully completed (p-value < 0.05) and a sample
was being compared with the reference, the hidden reference in the
ABC/hr window for that sample will be unhid.
You are about to answer a thread that has been inactive for more than 6 months. If you still wish to proceed, please ensure that your posting is original and does not duplicate or overlap any prior responses to this thread.