Data step processing

[SamuelRajKandru]

Is there any difference in the terms RECORD and OBSERVATION? Because the book says Input buffer holds a RECORD and PDV is the place where SAS holds an OBSERVATION. I'm thinking both the terms mean the same. Please help me if I'm wrong.

[RW9]

It seems to be quite clearly stated in the book there.  It referes to a "record" only when importing from a raw file.  The reason is because one observation may be split over many rows, you could read in row after row and then finally when you have all the information output the observation.  Take this example:

Name=abcde, surname=ghty

addr_line1=12 somewhere

addr_line2=earth

Now that is all one observation, but appears in 3 input records and to read it, you would hold the pointer over the line and continue until you had read the final data element.

[Astounding]

Going by my experience (not documented anywhere), there is no difference.  SAS tends to use the word OBSERVATION when talking about a record within a SAS data set, and RECORD when talking about a row of data from any other type of data.

While it's true that many raw data records can be combined into a single SAS observation, many other possibilities exist.  Just to list a few:

• Creating multiple observations in a SAS data set from a single record of raw data
• Creating multiple observations in an output SAS data set from a single observation within an incoming SAS data set
• Combining multiple observations from a SAS data set into a single observation in an output SAS data set

If you were to be specific about the type of data (SAS data set vs. raw data vs. some other type of data), and whether the file you are talking about is an input file or an output file, there would be no difference between RECORD and OBSERVATION.