Suggested Directory Structure for a Marketing Department:


/Product/Campaign/Datatype/Campaign_RawTable.Datatype

For a 'Fab' Campaign for Popcorn poppers:

Consider

/Pop/Fab/Sd1/Fab_Dem.sd1  /* Demographics of Target Population      */
            /Fab_Zip.sd1  /* Zipcode Characteristics of Target Pop  */

        /Sas/Fab_Dem.sas  /* Programs Log and Lst */
        /Rtf/Fab_Dem.rtf  /* rtf output           */
        /Cgm/Fab_Zip.cgm


Note file Fab_Dem.rtf ==> /Pop/Fab/Rtf ( but Fab must be unique tp Pop - Fab can have only one Father (Pop)  )


If I email Fab_RawDem.rtf to a user ouside my department, the user
could call me back weeks later and I would know exactly where
Fab_RawDem.rtf came from. Note 'Fab' should be a globally unique and
only associated with a specific Popcorn popper campaign.
'Fab' cannot be reused and should be registered.

There are analogs to this structure in Finance and pharmaceuticals.

Consider the SAS Tables

    Fab_Dem   /* users can add anthing after Dem, however I suggest TLAs */

       Suggested Variable Names for table Fab_Dem

         Fab_DemSsn  - Primary  Key Social Security Number
         Fab_DemSex  - Gender of Ssn
         Fab_DemAge  - Age of Ssn holder
         Fab_DemPay  - Income
         Fab_DemPet  - Ssn have a Pet Y/N
         Fab_DemZip
         etc

Develop your own TLAs and stick to them. Your programs will become more
readable and consistent from one project to another.

     Fab_Zip

         Fab_ZipZip  - Zipcode and Primary Key
         Fab_ZipPay  - Average Income in Zip
         Fab_ZipAge  - Ave Age of Head of Household

For a more advanced discussion of a naming convention in a relational
setting see soap001,2 and 7 at members.aol.com/xlr82sas/utl.html. I show an
example of a naming convention for a transper table, and with longer format names the
foreign keys in a transfer table can be set up as a format.

Why this structure:

      1. Clean SQL and Datastep Code - do not have to worry about
         name colisions. ( although you may need to rename key variable before a merge)

      2. When transposed the Names contain useful data, which can be parsed and
         has the potential to be powerful tokens. Very hady when using metadata sashelp.vcolumn.

      3. Self documenting code with improved debugging
         capabilities.

      4. Ability to trace the variables and tables through
         a complex multistep analysis

      5. Almost mandatory for full outer joins.
         Just consider a full join on four tables with
         many common variable names.

      6. Provides a structure for data normalization.

      7. Facilitates the implementation of relational transfer tables,
         these tables help resolve many to many joins.


/*--------------------------*\
| Analysis Datasets          |
\*--------------------------*/

 Analysis and composite Table Names and Table column names.

 Suppose we want to Summarize Fab_Dem at the Zip level.

 Possible new analysis  table name:

 Fab_DemSum  or Fab_DemZipSum

  Column Names

    Fab_DemZip
    Fab_DemAgeAvg  /* Average Age       */
    Fab_DemPayAvg
    Fab_DemPayMin
    Fab_DemPayMax
    Fab_DemPayStd
    Fab_DemFemCnt  /* Count of Fenmales */
    Fab_DemMalCnt  /* Count of Males    */

If you transpose this data and  can easily select
out the Pay statistics using Substr(name,8,3)='Pay'.

Note, if you stick with TLAs, you will tend to be
more consistent with all your programs, for instance always
use

Min - for minimum valyes
Max for Maximums
Ave for Average Values

Also consider putting the statistic TLAs last(Ave, Max, Min, 1st, 2nd etc ).

ie Fab_DemPayMax not Fab_DemMaxPay, Max is a property of Dem not Max.

etc.


Other tips on Variable Names, Campaign names etc

Suppose you already have a campaign number 2389 and
you need a TLA for this campaign.

4011 is  'FAB' in hex so you can use Fab as a TLA for
that campaign. You can even get 4011 back using sas informats.


Other SAS Variable names

Suppose Fab_DemSsn is character '123456789' and you
want to create numeric equivalent. Prefix the variable
with a lower case n. (c for character )

  nFab_DemSsn = Input(Fab_DemSsn,10.);

  or Fab_DemSsnN

  tFab_DemSav ==> temporary variable  or
  Fab_DemSavT

  lFab_DemSav ==> Local Macro Variable

  or Fab_DemSavL

  gFab_DemSav ==> Global Macro Variable

  or Fab_DemSavG


Other Prefix and Suffix possibilities:

   Foreign Key Variables in Dem Table

    Dem_Zip_ZipCde    ==> Column in the Dem table that is a foreign key to Zip



Useful Deviations

    Fab_FunWthSasForAllPeoPle

 If you cannot come up with a TLA, then string to TLAs together.
 ie PeoPle. This helps when parsing out for meta data.  Use PeoPle everwhere.

A  key to some good TLAs is two consonants with a vowel in the middle ie

Consonant ( a e i o u y ) Consonant

  Consider DayTym instead of DayTime
           NytTym instead of night time
           Fyl instead of File
           Byt instead of Byte

  Big, Max
  Ltl, Min

  Certain consonants sound like vowels

  B - E
  T - E
  C - E
  D - E
  L - A
  P - E
  V - E
  Z - E

  Zro -- you do not need a 'e' next to the Z
  NonZro
  GteZro - Greater than or equal to zero

  Txt - Text
  Btr - Better ( do not need the e's )
  Lbl - Label
  Lte - Late
  Lst - Last
  Lbr - Labor

---------------------------------------------------------------------------------

Slo - Slow
Flo - Flow
Gro - Grow

---------------------------------------------------------------------------------


Some numbers can help

  Ssn1st - first SSN
  Ssn2nd - Second SSN
  etc

If you are stll using 8.2 the number 2 can be used for formats

  Ssn2Sex  - SSN number to Sex
  Ssn2Wgt
  Ssn2Age

  Sex2Txt  ( 1- Female 2-Male )

-----------------------------------------------------------------------------------


 SAS Labels for variable names and SAS table names


 Proc Sql;

   Create
       Table Fab.Fab_DemSum
             (Label="Created by Fab_DemSum 1280 Rows 17 Columns Fab_Dem Summarized to Zip Level") as
   Select
        Fab_DemZip,
        Mean(Fab_DemAge)  as  Fab_DemAgeAve Label="Mean Age at Zipcode Level",

   .....