Return Transfers to the Original Format
Previous Topic  Next Topic 

Because systems vary widely in the types of variables they support and the information they store about them, you must be careful when transferring data from one format to another and then transferring back to the original format.  Stat/Transfer will always transfer numeric values without losing accuracy. However, other information about variables may be lost.


For example, SPSS maintains value labels and up to three types of missing value for each variable.  If you transfer from an SPSS data set to Paradox, the value labels will be dropped and missing values will be translated into a single Paradox internal format.  If you were then to transfer your data back to SPSS, missing numbers would all be translated into the SPSS system missing value, not the original values, and because Paradox does not store value labels in the data set, the value labels would be lost.


Because many systems do not explicitly support dates, you should be particularly cautious when planning two-way transfers of date variables.  Stat/Transfer will generally convert date variables into MM/DD/YY strings when transferring to systems that do not explicitly support dates.  This makes the date intelligible in the destination system.  However, when you transfer back to the source system, Stat/Transfer has no way of knowing that these strings are dates and will treat them simply as character variables.


If you are planning such a two-way transfer, you may wish to treat date data as numbers rather than as dates.  Stat/Transfer stores dates internally as the number of days since December 30, 1899.  Thus, if you wish, you can choose to put dates directly into your destination data set as numbers in this form by selecting the proper target type at the Variable Selection dialog box.  If you do so, these will survive a two-way transfer involving a system that does not explicitly support dates.