When a SAS/STAT procedure produces an output data set that contains observational-level data, you cannot assume that the order of observations (or rows) in the output data set and the original data set is the same if one of the data sets is (or both are) a CAS table. In this situation, the mapping of observations between the data sets might not be clear.
For example, the STDIZE procedure standardizes variables by transforming the original (unstandardized) observations in the DATA= data set into standardized observations in the OUT= data set. By default, the OUT= data set contains only the standardized variables. If you use a CAS table as either the DATA= or OUT= data set (or both), the observations in the DATA= and OUT= data sets might not be mapped in a predictable way. Sometimes this might not be an issue, because such a mapping of observations has no effect on subsequent analyses. Problems might arise when you must establish a correspondence of observations between the data sets.
Table 15 summarizes known mapping issues of the observations in the OUT= data set when you use a CAS table as the input (or output, in the case of PROC STDIZE) data set. In these cases, SAS/STAT procedures issue a note to the log and recommend remedial actions.
Table 15: Mapping Issues of the Observations in the OUT= Data Set
| Procedure | Issue | Solution |
|---|---|---|
| CLUSTER | 'OB1', 'OB2', and so on, as the _NAME_ variable values cannot be used to identify the observations | Specify ID variables |
| MODECLUS | The _OBS_ variable values cannot be used to identify observations | Specify ID variables |
| PRINQUAL | 'ROW1', 'ROW2', and so on, as the _NAME_ variable values cannot be used to identify the observations | Specify ID variables |
| STDIZE | No predictable mapping between the DATA= and OUT= observations (also applicable when the OUT= data set is a CAS table) | Use the OPREFIX= option to ensure unstandardized and standardized variables are both available in the OUT= data set |