Table of Contents
avg - ERP raw data signal averager
avg presampling avgfile
avg performs signal averaging on ERP raw data. Three
files must be specified after the program has been invoked: a raw data
file, a log file (both obtained from the dig(2)
program), and a binlist
file (most often obtained using the cdbl(1)
program). More than one set
of raw, log, and binlist files may be averaged together. Presampling specifies
the length in milliseconds of the pre-stimulus portion of the epoch; it
cannot exceed 256 times the sampling period (at the sampling rate used
during digitization) and must be greater than or equal to the onset delay
specified at digitization time. Note that presampling represents the total
prestimulus period desired, including the onset delay (from trigger to
stimulus). The averaged data are written to avgfile.
The options are:
- -a arfile
- Perform artifact rejection using the tests specified in arfile.
For more information on artifact rejection and the format of this file,
see the garv(1)
- Mark the associated log file during averaging
to denote events which were not included in the averages due to data error,
and due to artifacts if the -a option is specified. These events are marked
by setting the log flag 0100 (octal) for data errors, 040 (octal) for events
rejected by any of the artifact rejection tests, and 020 (octal) for events
rejected specifically by the polinv polarity inversion test (note that
for polinv rejected events, both the 040 and 020 flags are set, which for
example results in 60 being displayed in the ’Flags’ column of the logexam(1)
Artifacts rejected by the polarity inversion test (i.e. blinks)
are given special treatment because they are the only type of artifact
that can be corrected. When avg runs with the -a option, events with polinv
artifacts are excluded from the bin averages along with events containing
other artifacts as specified by the tests in the artifact rejection file.
However, if avg is also run with the -b option, the blinks can be corrected
later with blinkcp(1)
. In this case, the polinv rejected events will be
corrected and included in the corrected bin average. Giving polinv artifacts
a unique log flag allows these blink events to be identified. This can
be useful, for instance, if one wishes to compute behavioral measures based
only on EEG events that were not rejected. Before applying blinkcp(1)
polinv events are rejected, after blinkcp(1)
they are not: the polinv log
flag allows you to keep track of these events.
When the -x option is specified,
the log file is processed to clear the above flag bits prior to averaging.
If the -x option is not used, the log file is read only to ascertain that
the data file, logfile, and binlist files belong together and to detect
mismatches. In this case the log entries corresponding to the events that
are rejected remain as they were, and the bits are not cleared prior to
Note that only events selected for averaging are tested for
data error or artifact rejection, so when using the -x option only the events
rejected from this set are marked in the log file.
- -r dec_factor
the sampling rate of the data (decimate) by an integral factor of dec_factor
as they are extracted from the raw file.
This is one way to enable the
average of long epochs, but be aware of the resulting loss of resolution
in the waveform - specifying a dec_factor of 2 will effectively reduce 250Hz
sampled data to 125Hz sampled data.
Also see the -c option for enabling
the average of long epochs.
- -c chan_prec
- Instead of employing 256 points
in each averaged epoch, use chan_prec * 256, thus increasing the epoch
length (as well as the amount of space needed).
This is one way to enable
the average of long epochs, but be aware of the increase in the size of
the average file - doubling the size of the epoch will pretty much double
the size of the average file.
Also see the -r option for enabling the average
of long epochs.
- Calculate standard error and store it as a second dataset,
or processing function, in avgfile. The ploterps_manual(7)
gives an example
of how to access and plot the standard error data from this second dataset
created in avgfile. The ability of erp average files to store multiple
datasets, or processing functions, was never really exploited; this option
may be the only use of this facility in the erp programs.
- -o outfile
stats such as mismatches to outfile as well as to the screen.
- -b blinkfile
- Create the file blinkfile for subsequent use by the blinkcp(1)
This file contains averages for the epochs that were determined to contain
blink artifacts by the polinv artifact rejection test (see garv(1)
using the -b option, you must also use the -a option, AND the artifact rejection
file must invoke the polinv artifact rejection test.
- Work quietly.
filt_type cutoff [cutoff]
- Filter the data (before artifact rejection),
where filt_type is one of lp (lowpass), hp (hipass), or bp (bandpass). The
lowpass and hipass filters require a cutoff frequency to be specified;
the bandpass filter requires both low cutoff and high cutoff frequencies
to be specified. Use of this option is generally discouraged, but it
can be useful in special cases such as cleaning data contaminated by 60Hz
noise. Also, it is recommended that the resulting average file name be
chosen so that it somehow reflects the fact that the data contained in
it are filtered.
If all is well following invocation, avg will print out:
Enter raw, log, and bin list filenames...
As long as there remain files
to be averaged, specify them here. When averaging is complete, a summary
of the various bins and certain data parameters is printed on the standard
Too many too list here, most
are explained in the "avg User’s Manual".
Table of Contents