may be as simple as 1 for delete, 2 for add, and 3 for replace. The simple actions of deleting duplicate or completely erroneous records and adding new records or supplying omissions from the original file are easily accomplished by a code 1 or 2. The changing of items other than the location and identification codes requires only one card with a code 3. For changes in the location identification codes, one card (code 1) will be needed to delete the old record and one card (code 2) to add the new record.

2.2 Hand counts

The hand counts are the DPD Control Clerk's counts of the number of establishments for each ED by type of form. The counts are posted to the Master Control Sheet (exhibit V-1-2 or V-1-3). These counts are the control factors throughout processing.

The hand counts may contain some errors. To verify the count before tabulation is not usually worth the time and effort. The best way to verify the hand count is to compare it with the number of establishments punched for the ED. This comparison will be made in Providencia after the cards for an ED are punched and the data are put on tape; significant differences are investigated and corrections are made on the tape as necessary. In the meantime, the hand counts are accurate enough to serve as estimates of production for the editing-coding and key punching operations. 2.3 Magnetic tape filing control

One very important part of a control system is the establishment and maintenance

of a filing system for the computer tapes. Since several hundred tapes will be required to adequately handle the full census program, there may be filing difficulties if the supply of tapes is not sufficient or if they are not adequately controlled.

Occasionally, information on a magnetic tape is erroneously erased. This can be caused accidentally by using the tape at the wrong time or (as happens only rarely) the tape may be exposed to a strong magnetic force that will erase or change the magnetic images on the tape. To provide for this contingency, backup tapes should be kept and filed in the tape library. Backup tapes are copies of the tapes used for processing. They are normally kept for one phase of the processing until the next phase or two have been completed, reviewed, and corrected as necessary. Then if data need to be re-created, only one step is needed to bring the tape to the desired or current

status.

When the computer staff keeps a backup tape for each batch processed, the number of tapes that are needed can multiply very rapidly. Of course, all tapes can be re-used eventually, so the cost is not a waste. In calculating the budget, however, the initial cost of extra tapes should be included.

Because a large number of tapes are needed for the census, it is essential that ample space be provided for setting up a tape library. It should be close to the computer and away from large motors, generators, or equipment that may create a strong magnetic force. The library should be equipped with multiple sections of shelves designed specifically to hold magnetic tapes. The number of shelves and the amount of space needed for the shelves must be provided for in the plan for filing the tapes. If some of the shelves are free standing, allowance must be made for passageways wide enough for a cart or other carrier. If the sections are labeled and each tape position (or slot) within the section is

3

numbered, than a record of the location of

any one tape can be maintained. To facilitate that record, definite sections should be assigned to each major area of processing. Under no circumstances should blank or work tapes be stored with good data tapes.

The control and tape storage specialist, or simply tape librarian, is responsible for setting up the library, assigning the sections, filing and maintaining records on individual tape locations, insuring that the proper tape I is used for each step in the processing and tapes are not lost or damaged, verifying that the proper tapes are blanked for re-use upon request, and seeing that unauthorized persons do not have access to tapes.

2.31 External labels.--Another valuable aid in tape control is a good system of tape labeling, both internal and external. External labels are visible labels, physically attached to the reel of tape. They contain information such as file number or name, reel sequence number, date of run, description of the contents of the reel, and the library location. As soon as a tape reel has been created, the "file protect" ring should be removed in order to prevent accidentally writing over the good information.

2.4 Card-to-tape control

Control during the card-to-tape process provides for the computer to compare the hand counts (which were punched in the ED cards) with totals tabulated from the actual punched data. To aid in resolving any differences between the hand counts and the data as punched, the card-to-tape diary lists the individual establishments, the hand counts of establishments and card types, the tabulated counts of establishments and card types, and the percentage of variation by ED and work unit.

A diary is a print-out of certain information which is important in the analysis of the procedure. The diaries are examined by subject-matter specialists, who determine what action, if any, is to be taken. An illustration of a card-to-tape diary is found in exhibit V-6-2.

2.5 Edit control

Control during the machine edits is provided by review of the edit diary. Tabulated counts similar to those in the card-to-tape diary by ED and work unit, and the number of imputations by item, are provided for review. In the diary, records with a high imputation rate can be shown in two forms: (a) the record as punched and (b) the record with the 'imputations; this would reassure the subject-matter specialists that the imputations have been made in an acceptable manner where required. For example, particular items from the questionnaire

will be zero balanced during this edit procedure; that is, totals for specific items must be the sum of the individual fields in that item. The edit diary would essentially act as a zero balance listing in this instance.

When the tabulated counts appear not to balance, the card-to-tape diary should be checked to assure that the latest or accepted version is being used, and that the tape input was the accepted or corrected tape. An investigation of a high number of imputations may disclose that: (a) the instructions need to be revised or personnel should be re-trained at one or more points in the processing, or (b) the problem occurred during the enumeration process and the only possible action is to review the imputation process.

2.6 Tabulated control counts

The Master ED File print-out and diary serve a very useful purpose. They provide tabulated counts for the various geographic areas for use as control totals during the tabulation process. For each tabulation thereafter, the Master ED File print-out gives the final counts for each ED.

2.7 Filing punched cards

At times during the machine edit and the diary reviews, it is necessary to refer back to the punched cards. Provision should be made to file the materials so that they can be located readily. Paper and boxes of punched cards are heavy. Shelves should be of material sturdy enough to carry the weight. An estimate of the number of boxes of cards provides the basis for estimating the amount of shelving and the area for passageways. For example, the standard size box for punch cards is approximately 37 by 19 by 9 centimeters (14 1/2 by 7 1/2 by 3 1/2 inches) and holds approximately 2,000 cards. Provision must be made for accommodating such equipment.

3

The card designs that were described earlier for the most part included all the data to be punched from the questionnaires and the necessary geographic identification. However, there are some additional items that are needed in the tabulation process, and there needs to be some re-arrangement of the record formats. It is wise to consider the total requirements when the layout is first planned so that the record will be long enough to accommodate all the items that will be needed. The following is a list of items (codes) that are needed and the processing step in which the codes can be added to the record:

additional effort to keep in mind the various formats and sizes. The decision needs to be made to completely re-format during the cardto-tape run, allowing space for codes that would be needed but would not be available until later.

3.4 Tape blocking

Before the computer can write data on a tape, the tape must be moving past the read/ write head at the required speed. The read/ write head is the mechanism which actually reads or writes the data.

After the write, the tape stops moving. The same is true when reading data. The amount of tape that is passed during the speed-up and slow-down is called an interrecord gap and is approximately 1 1/2 to 2 centimeters (0.6 to 0.75 inches). Since one record for the various card types requires approximately 0.2 centimeters (0.08 inches), most of the tape would be blank if the record for each establishment were to be written as a separate record on tape. Also, the start and stop time would be greater than the actual read or write time. For efficiency, the normal procedure is to write several records as one physical tape record. This procedure is called blocking and the physical tape record is called a "block."

There are several factors to be considered when determining how many records should be combined to produce an efficient block size. The major factor is the amount of computer memory that will be available to store the block throughout the entire processing system.

If all records are of the same size, it becomes a matter of determining how many records will fit evenly into the space available. For example, if the maximum amount of computer memory available for tape input and/ or output (read/write) throughout all the

processing steps is 1,000 positions, the blocking factor would be 12 records per block. With the use of blocking factors generally, space would be well utilized and total data access time would be decreased.

3.5 Card-to-tape diary

At the time of the card-to-tape run, a diary or printed list will be prepared. Like any diary, it tells what actually happened. during the processing for each ED. The diary serves as a control, a record of processing, and an aid in balancing the ED totals with the totals as tabulated from the punched cards.

3.51 Content and preparation of diary.-The cards are read and checked for errors as described earlier under validity tests. When an error is detected, an appropriate error code and the identification of the card in error are printed in addition to the tabular portion shown in exhibit V-6-1. Upon completion of the work unit, the total establishments and card types from the work unit card, the tabulated counts, the percent of variation between the two, and a total record count are printed. Each work unit will start on a new page and all pages will be clearly identified as in the example and numbered. As an aid to balancing the totals, the serial number of each establishment is listed. The serial number makes it easy to locate the discrepancy.

3.52 Review of the diary. --The diary is reviewed for each Work Unit. The reviewer should be a subject-matter person who would consult (when necessary) with a data-processing specialist and would recommend what action should be taken. The reviewer should be assisted by a clerk, who could look into many of the discrepancies. Errors in geographic identification codes will likely cause an entire ED to be rejected for re-punching.

The most likely discrepancy will be a variation between the counts (as reported by the Control Operations Branch) and the actual count of establishments and cards that have been punched. It is expected that there will be some work units for which the card count on the tape will be higher and some will be lower than the counts reported by the DPD Control Clerk. It is assumed that the quality control procedures affecting manual editing-coding and card punching will have minimized discrepancies by the time the cards are put on tape. To check punch cards one by one with the Control Clerk's entries would be a time-consuming and costly operation--out of proportion to its value. Simple manual checks of this type are also subject to considerable

error.

In Providencia, tolerances were set up to permit small differences between the Control Clerk's count and the count of cards on the tape. If the two counts differed by more than 2%, the reviewer would investigate the discrepancy and make whatever corrections are required. As the reviewer gains experience, he will learn where to look for likely errors. 3.53 Corrective action. --Some programming routines are already available which provide for correction of data already in a tape file. A correction routine provides for additions and deletions to the file. For processing a census which is taken infrequently, the time and cost to write the correction routine may be greater than the clerical cost to correct the work unit. It may be more feasible to re-punch the cards for a rejected work unit and then re-process that unit.

An alternative procedure would reduce the total amount of processing but would add to the task of controlling the movement of materials. A list of rejected ED's by work