|
|
Docs: ERPSS System Tour - Page 5
Normalization of data files is a mandatory procedure that must be performed on averaged data files prior to using them with other ERPSS programs. Once data have been normalized, they can be measured, manipulated, and plotted as described in subsequent sections. Normalization standardizes ERPSS data files so that numerical values represent definite voltages, since the gain of recording apparatus can vary with experimental setups as well as between channels. This normalization is performed by measuring the size of a known calibration pulse separately for each channel, and then scaling each data channel, given the desired data resolution. Bin 0 is reserved for this calibration data in averaged data files.
The program normerps (E1) performs normalization. It is usually used interactively along with a plotting device, so that one can verify that accurate measurements are being made on the calibration data so that normalization will be accurate. However, it can be used non-interactively, if desired. When invoked interactively, normerps plots the specified calibration data, and then interactively allows the user to vary parameters that affect the normalization process. The size of the calibration pulse is measured using two cursors, propitiously positioned for accuracy. One can specify a number of points on both sides of each cursor that will be averaged, alter the plot magnification, set the desired output resolution (in pts/uV), specify the polarity of the data, select subsets of channels to display, quit immediately without normalization, or proceed to normalize using the current parameters.
A more tutorial user's manual, however, is also available which includes a detailed discussion of the issues and considerations involved when data are normalized.
Since ERPSS data files are stored in binary form along with interspersed header information, they cannot be "edited" or altered using standard utilities. Hence, the formation of difference waves, grand averages, and the like require special programs. This section describes those programs and their function in the ERP software system.
On occasion one may need to delete certain bins from a data file, delete certain channels from a data file, combine bins from different data files, and the like. These functions can be accomplished using the program datacat (E1). Note that the resolution (determined during normalization) must be the same for all files the bins from which are to be combined with datacat.
Usually one will need to further process ERP data after averaging and normalization before plotting and statistical analysis. This is done using erpmanip (E1). Erpmanip can perform arbitrary linear operations on event-related potentials (ERPs) in standard ERPSS data files. It uses an ASCII command file, termed a erpmanip command file (ecf), to specify various operations on the input data, and places the results in a new output file. Any linear combination of channels and/or bins can be generated using erpmanip, including scaling by the number of sums. Erpmanip can be used to calculate difference waveforms, averages across stimuli or conditions, "bipolar" derivations, re-referenced waveforms, or current source density (CSD) estimates.
Generally one records and analyzes data from a number of subjects, processing it similarly. The program grandavg (E1) is used to average these similar files, bin by bin, forming a grand average across all subjects. Although this is its intended function, grandavg can be used to average together any set of data files, bin by bin, as long as they are compatible. The manual page (hopefully) says it all.
It is possible to digitally filter averaged ERPSS data files using the program idfilter (E1). Digital filtering of the data entails convolving the data samples with the impulse response of the filter. This process can be described in the frequency (Fourier) domain as a multiplication of the spectrum of the data by the frequency response of the filter. However, since the convolution can require additional points at the ends of the un-filtered data segment in order to produce a certain number of filtered points, one must anticipate this "fringe loss" and average additional points at the beginning and end of the data epochs.
The actual filtering of data using idfilter is simple; the real decisions involve the selection of the filter. To this end, a number of support programs have been provided to assist in this endeavor.
Although it does not correspond to any particular ERPSS program, there is a manual entry in section one of the ERPSS manual entitled "idf" which contains an introduction to the issues, considerations, and operations involved in filtering ERPSS data using integer digital filters (idfs). This is the place to begin; the document discusses the functions and use of the ERPSS idf filtering programs. Below follows, however, a brief summary of the auxilliary support programs that one may wish to use in the process of selecting a filter.
Although there are a number of pre-constructed filters that are available in the lib/idf subdirectory of the ERPSS file system hierarchy, a number of programs are available to construct filters, if necessary, to attain a certain response. See makeidf (E1) for more information.
When one wishes to identify an unknown filter or print the impulse response for a particular idf, one uses printidf (E1).
This program is the primary resource available to assist in the selection of an idf. It allows one to view the response of an idf in both the time and frequency domains, displays the result of applying the filter to real or simulated ERP data, and determines the number of points that will be lost due to "fringe effects". It operates interactively, and is well described in the manual entry plotidf (E1).
Here is a sample of the output from plotidf. The input idf file was /epl/erpss/lib/idf/gauss.l.7.5:
This program actually applies a specified filter to a set of ERP waveforms. The filter is applied to all of the channels in all of the bins in a data file. The plotidf program is typically used first to select an appropriate filter, and idfilter (E1) is then used to filter the data contained in a file.
Measuring various parameters of ERP waveforms is an essential step in a statistical analysis of an experiment. The program merps (E1) is available for performing these measurements, and produces a stream of ASCII data values corresponding to a sequence of specified measurements. It employs one or more measurement command files (mcfs), standard text files that are created using an editor, to control the type of measures, the order in which they are performed, and the data on which they are made. When used to perform measurements that will be used in subsequent statistical analyses, merps is used in a non-interactive mode, in which only the measurements themselves, without any descriptive information, are placed in an output file. However, it is often useful to more closely monitor the measurement process to ensure that the proper data are being measured in the proper order. For this purpose, merps is invoked interactively, usually in conjunction with a graphics device. In interactive mode, merps prints additional information about each measurement, and the data being measured can be displayed on the graphics device, if desired.
The details of the use of merps are contained in a user's manual; that document describes the considerations involved in performing measurements, the measurement functions that are available, the format of an mcf, etc. One note - the order in which measurements are made is of the utmost importance. It must correspond to the order in which data are required by the subsequent analysis program (e.g. ranova (E1) ), or spurious results will be obtained. There is no built-in system for verifying that the order in which data were measured corresponds to that specified in the analysis program - care should be execised if veridical results are desired.
At times it is useful to print ERP waveforms as a series of ASCII values in order to transfer the data to another machine or to a program that accepts data in this format. The program erptoasc (E1) can be used to accomplish this conversion. ASCII data files can also be converted into ERPSS-format data files by using a complementary program, asctoerp (E1).