A record is a group of related data items. In the process of creating a record to define (a portion of) the questionnaire, you will also be defining the physical layout of the data file. For example, suppose your (very simple) population record looks like the following (only item name, starting position, and length properties are shown; starting positions show that ID items occupy the first 9 positions in the record):
Item Name Start Pos Length
RELATIONSHIP 10 1
SEX 11 1
AGE 12 2
If an operator had keyed a questionnaire for a 35-year-old female (Sex = 2) head of household (Relationship = 1), you would see a line in the data file, corresponding to the population record defined above:
1235 (line in data file)
In deciding on a file structure, there is often the choice of defining a record type which occurs once within a questionnaire but contains repeating sets of data, or to define a record type which occurs multiple times within a questionnaire, each with a single occurrence of the data. The application designer should take into consideration the amount of information that recurs and the probable number of occurrences.
A common example in a Housing and Population Census is information about deaths in the household during the 12 months prior to the census. If this information (usually sex and age at death of the deceased) is collected during enumeration, the expectation is that 95% of households will have no more than one, or at most two, deaths during the previous 12 months. With this volume of information, it would be practical to have one record type that occurs once within the questionnaire and allows for repeating occurrences of the data, since it is unlikely that even the maximum number of occurrences, multiplied by the number of positions occupied by each occurrence, will exceed the length of the already-existing household and population records.
However, in the case of an agricultural survey, a section on crops may include questions about acreage planted, yields, etc., whose cumulative length for each crop mentioned may be quite large in relation to other records in the file. In such a case, it would be more practical to define a record type that occurs multiple times within the questionnaire. Within each occurrence of the record would be found the information relating to one specific crop.