Tom You are right to aim at some basics. It are the concepts to be reviewed: SAS 9.3 Programmer's Bookshelf Some nasty things: 1/ Missings For statisticians it it quite common and makes sense to have many of them with different meanings. - The SQL concept is failing in this part, just having one value "missing". - SAS SPSS Stata (and others) are solving that it in different ways but supporting that. SAS has the option to work with many types: SAS(R) 9.3 Statements: Reference (missing). - The 9 habit as used in surveys is something going back to the hollerith approach, just having the numbers 0-9 as options. That time has passed but habits are pertinacious. At the data-cleansing step this should solved, adjusted an old proceedng http://www.nesug.org/proceedings/nesug01/ps/ps8009.pdf And (stata) Stata FAQ: How can I recode missing values into different categories? 2/ Formats Nothing is what is what is seems to be. The numeric (floating) technical handling is rather basic as some quirks can pop-up. The rounding and difference with integers is a pitfall. Dates Times and datetimes are challenging. Unicode for the next frontier. The concept of formats is rather unique with SAS. It is the only one environment/language I have seen where this is a segregated part of the data and can be seperately maintained. This offers quite smart and simple approaches not well known. Base SAS(R) 9.3 Procedures Guide, Second Edition (proc format). You could use MLF (Multi Labe Format) own proceduers (FCMP) The switch in thinking: No need to adjust data generate additonal datasets, just apply different formats at analyzing. Note: Base SAS(R) 9.3 Procedures Guide, Second Edition (Proc report) This procedure has a compute block (it is exceptional)
... View more
