To process the data correctly, some procedures require that an input (or output) data set be stored as a special data set (rather than as a raw data set). Such a special data set might rely on the order of observations to preserve some implicit information about the data. For example, when reading a TYPE=DISTANCE data set, PROC CLUSTER assumes that the observations (rows) in the data set are ordered in the same way as the objects that are represented as variables (columns). However, because a definite order of the observations in CAS tables cannot be assumed, PROC CLUSTER issues an error when you attempt to use a CAS table as the DATA= data set that is a TYPE=DISTANCE data set. In this case, you must instead input a TYPE=DISTANCE SAS data set that has the required or intended order of observations.
For more information about incompatible special data sets, see the section Special Data Sets Incompatible with CAS Tables of Appendix C, Known Issues in Using CAS Tables with SAS/STAT Procedures.