The data is being stored at the University of California, Irvine, at http://kdd.ics.uci.edu/databases/eeg/eeg.html.
Acknowledgments for donating space for storing the large archive should
be made to
Bay, S. D. (1999). The UCI KDD Archive [http://kdd.ics.uci.edu]. Irvine, CA: University of California, Department of Information and Computer Science.
Tables of ASA-fitted parameters and 60 PostScript figures, each figure
containing 4 6-electrode graphs of potential and CMI for testing and
training data sets, for each of the 10 subjects for each of 3 experimental
paradigms, are in
I am not an expert in the clinical aspects of the alcoholism studies which generated this data, and therefore cannot offer advice on these clinical paradigms. For many studies, it will suffice that the data is clean raw EEG, and this should be useful to many people.
All data is compressed in gzip (.gz) format, resulting in a reduction of size of files by about a factor of four. See utils_file_formats.txt for some links to information on gzip, PostScript, PDF, tar, and shar utilties.
The data used in smni97_cmi.ps.gz for the 3 matching paradigms, c_1 (one
presentation only), c_m (match to previous presentation) and
c_n (no-match to previous presentation), for the 2 subjects,
alcoholic a_co2a0000364 and control c_co2c0000337, 10 runs each,
may be downloaded as a tar'd directory containing 6 files,
The complete data set used as described smni97_cmi.ps.gz, 10 alcoholic and 10 control subjects, 10 runs per subject per paradigm, was used to produce a set of training results.
A follow-up study described in smni97_cmi_test.ps.gz used the same 10 alcoholic and 10 control subjects, but with 10 out-of-sample runs per subject per paradigm (using a small percentage of training data to fill out 10 runs per subject per paradigm for 6 of the 60 cases), producing a set of test results.
This data can be downloaded as
Examine the header in each file to determine under which paradigm the run took place.
Each of the 11,075 typical gzip'd files in the 122 tar'd directories is about 65K. The entire set of data is about 700 MBytes. For example, a single file obtained from expanding tar'd directory co2c0000337.tar.gz would be co2c0000337/co2c0000337.rd.047.gz, where coc... represents a control subject and coa... represents an alcoholic subject. Examine the header in each file to determine under which paradigm the run took place. All these classifications for purposes of calculation are easily automated.
NOTE: There are 17 trials with empty files in co2c1000367. Some trials have "err" notices, e.g., search/grep for "err" and see "S2 match err" or "S2 nomatch err" etc.
Access to all these tar'd directories is available from
The full dataset in one tarfile can be downloaded as
Lester Ingber <firstname.lastname@example.org> Copyright © 1997-2012 Lester Ingber. All Rights Reserved.
$Id: smni97_eeg_data.html,v 10.103 2012/08/07 15:12:41 ingber Exp ingber $