The audit function is critical to ensure the accuracy of data
analysis. CODAP has programs to audit all critical input data.
AUDITF, AUDITM, AUDITT and AUDITV check format, module title, task
title, and variable title cards respectively. These programs
should always be run before any CODAP analysis is performed.
The raw data file containing the survey responses is checked by AUDITR. AUDITR will read any text raw case data input file. Each case may consist of up to 200 records, each record containing 84 characters. The records must be in proper sequence as defined by the format cards. Each case may contain a maximum of 8000 characters of background data encoded into 2000 computer words. Note that a one digit background response and a four digit response both use one computer word each. The maximum number of task responses permitted is 3000.
Common data errors include too many or too few cards for a given case, too many or too few task responses (default minimum five, maximum 95 percent of the inventory), or check characters which do not match. AUDITR also range checks the task responses. The user may specify the valid task range (default is 1 to 9) on the AUDITR control card. Zeros and blanks are considered different unless the 'Z' option is specified. Task responses not in range are reset to a 'non-response.' Note that a 'non-response' is not equal to a zero response.
AUDITR creates two output files. The first file is the data file containing valid cases and the second is an error file with cases that had problems. If the errors are corrected (using the Text Editor), AUDITR can then be run again to merge the corrections onto the valid file.
The AUDITR listing, which consists of four different reports, should always be carefully reviewed. The first report lists all cases which were written to the error file, along with the reason for deletion. The second report displays all unspecified card columns which contained data. This can help detect possible errors in the format cards. The third report counts the number of invalid task responses for each card for each case. The final report displays interval frequency distributions of the total number of valid task responses for each case.
Once the audit programs have been completed, INPSTD may be run. The INPSTD program accepts the good output file produced by AUDITR and produces a Case Data file. INPSTD will accept either rater or incumbent data as input. Incumbent data responses are converted into relative time spent values. INPSTD also creates the first ten computed variables and three sets of duty variables (percent time spent in each duty, number of tasks performed in each duty, and percentage of tasks performed in each duty).
INPSTD can automatically generate a series of products to validate the correctness of the Case Data file, such as a TASKXX listing of the tasks, a DICTXX roster of background variables, a PRTJOB report of the job description for the entire sample, and an AUDITD distribution of all background variables which are six characters or less in length. If rater data is being used, the 'R' option should be specified. With this option, INPSTD can automatically generate a GRPREL run, which reports the interrater reliability for the total sample. The 'A' option will cause GRPREL to adjust and standardize the ratings (typically used for Task Learning Difficulty data).