This file contains two major kinds of summary data derived from Case Data files. This summary data includes membership (i.e., group) definitions, and task factor information. The membership information for each group is stored as a packed bit- mask. The bit-mask is a series of ones and zeroes which indicate membership or non-membership in the specified group. The number and sequence of these ones and zeroes is directly related to the cases on a Case Data file family. In other words, membership masks created for a job incumbent Case Data file will not work on a subject-matter-expert Case Data file, even if the two files contain many people in common.
Task factor information is stored as vectors which have one value for each task in the job inventory. The three major varieties of task factors are direct ratings, incidental measures, and hybrid factors. Direct ratings include job description information and subject-matter-expert ratings. Job description vectors include percent members performing (PMP) and percent time spent (PTS) by all members. Each job description is associated with a specific group and provides the context for the phrase "by all members." The percent time spent vector contains the average percentage of relative time spent ratings, including zeroes, for all members in the group for each task. The group's identity and number of members is recorded in a preamble to each task vector. As noted above, the group's identity can be decomposed into individual case identities by using the membership mask.
The second subtype of direct rating vector is created by evaluating and refining subject-matter-experts' ratings. These vectors are usually generated using the GRPREL program, which not only reports the interrater reliability, but also creates and stores tasks factors which represent the mean, standard deviation, and number of raters for each task. Examples of these factors include task learning difficulty and recommended training emphasis.
The second major variety of task factor is incidental measure. This covers factors which are created based on making computations incidental to the original purpose of the rating. For example, although job incumbents rate tasks on relative time spent, one can compute an average grade level for all people who have rated (i.e., perform) a given task. The GRPAVG program will compute this value for each task and store the resulting task factor. Other examples of incidental measures include task discrimination, task presentation sequence, task co-performance and module reference factors.
The third major variety of task factor is the hybrid factor. Hybrid task factors are created by processing other task factors. The simplest example is the RNKFAC program which creates a new factor by replacing the raw values of the source task factor with their corresponding rank order value based on either an ascending or descending sorted order. Another example is TASCAT, which sorts a continuous task factor and then creates subdivisions or modules based upon the ordinal positions. The created task factor contains the "category" or module number in which each task was allocated. The program can also create a Module Title file which can be used to report tasks in complex orders to assist in task screening or selection methodologies. The most versatile form of the hybrid factor is created by the FACGEN program. FACGEN can perform any mathematical computation on any task factor, including previously defined factors as well as factors in the process of being created. Since FACGEN has immediate access to the overall mean and standard deviation of any previously computed task factor, it can quite easily produce a standardized or normalized version of any factor.