Customer Service: +44 1334 657957

Split data records

This method may be used if a little QPSMR CL expertise is available.

The data is entered in the same order as it was collected.

It will help if extra space is allowed in the serial number width when the questionnaire defined.

The conversion will be much simpler if ASCII data is used rather than binary data.

A simple QPSMR CL recode is written to produce a new data file. This new data file will contain a record for each repeat.

The recode does the following:

  • Reads a questionnaire.
  • Changes the serial number by placing a 1 in an unused serial column. A 1 is also placed in an unused location at the end of the questionnaire. Then outputs a copy of the data. This will serve for respondent tables and the first repeat
  • Changes the serial number by placing a 2 in the serial column. A 2 is also placed in the location at the end of the questionnaire. Then copies all second repeat locations to overwrite the first repeat locations. Then outputs a copy of the data. This will serve for the second repeat.
  • Do this again for any further repeats. If more than 9 repeats are needed then two unused data locations will be needed in the serial number.
  • The QDF is then copied and changed to delete all second and subsequent repeats. A new question called REPEAT is added which picks up the extra repeat number from the end. This in turn is used to make a variable called LEVEL which has two responses:
  • LEVEL/1 is REPEAT/1 (respondent)
  • LEVEL/2 is T (true) (product)

When tables are run from the new data file with the new QDF all respondent based tables are filtered on LEVEL/1 and all product based tables are filtered on LEVEL/2. It is no longer necessary to produce a product table for each repeat because this is being done by feeding in a new questionnaire data record for each repeat.

The main problem with this method is that each questionnaire has been split up into separate records which cannot be cross-referenced. For example you cannot refer to any product on respondent based tables, nor can you refer to products other than the product being analysed in product based tables. This difficulty can be overcome by defining new variables which are then added to each record during the CL conversion.