DATA PROCESSING OF ANTI-DANDRUFF RESEARCH 791 Transmittal Processing After the study was completed, transmittals were edited for com- pleteness and accuracy. The doctor's comments on the record forms and the subject's recorded likes and dislikes on the questionnaires were coded at this time using master code lists. Next, the pertinent informa- tion was key punched into 80-column data cards. The cards were veri- fied to insure accuracy. Because there are only 80 columns per card, a separate card was required for each transmittal form. Unit Record Machine Processing The key punched cards were sorted to group number and subject number sequence and separated into three card types--history, record, and questionnaire--on a high-speed card sorter. Then each deck was listed on a tabulating machine for the purpose of visually editing the key punch data. After necessary corrections were made, the history deck was matched against the record and questionnaire decks on group and subject numbers to insure that all three cards for each patient had been key punched for processing. The computer was programmed to note in- consistencies or logic errors in the raw data. If a card for a patient was missing in any of the three decks or if an identifying number was incor- rectly key punched, this would constitute selected unmatched cards. Reasons for unmatched cards, other than the two mentioned above, could be as follows: (a) A subject voluntarily dropped out of the study. (b) A subject failed to follow the prescribed protocol instructions and was dropped. (c) A transmittal form was inadvertently not key punched. All unmatched cards ejected in the sort and match process were checked at this time, and necessary corrections were made before processing could continue (Fig. 6). The next phase in unit record processing was to segment the history cards into desired age groups and years with dandruff groups for correla- tion reports. This was accomplished by sorting the history cards into the age or years with dandruff sequence and separating the cards into the desired groups. Then each group was matched against the record cards on test product and subject number in order to separate the record deck into the same groups. The segmented record cards were then coded using the gang punching method on a reproducer. Thus, the assigned codes were used as control factors in computer processing rather than the

"792 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS UNIVAC P 2•480 Figure 4. Patient history data card Key Punched & Verified Data Cards Preliminary sort to test product and patient number sequence. Separate into card types. Tabulating •. 8•• Listing for Visual •achine Edit Match History deck against record and questionnaire decks. -- Unmatched Unmatched Correct errors and Cards Cards Rematch I Figure 5. Edit phase of unit record card processing