/************* for 1990 Census Files *********************************
NOTE: You can copy this to a file directly from here, and use 
      the program shown below as a sample, so that it need not be 
      re-typed entirely.  Depending on your browser, you can probably
      click 'File' then 'Save As' to save this into a usable file.  
**********************************************************************

Making Subsets of 1990 STF1A Census Data for Use in Packages 
Other than SAS or SPSS
----------------------------------------------------------------------

Some of the Census files are larger than 1 Gb, and as such,
often cannot be moved to UNIX disk.  This means that creating
a text-only (ASCII) subset of the data may be helpful.
SAS can be used for this purpose.  When the data you want are in a 
SAS data set, a simple SAS PUT statement can output the data into 
"raw" (ascii) form to be read into whatever package you are going 
to use.  A sample SAS program to read in a subset of Census variables 
from the STF1A file and output raw data is shown here.  (This
directory also contains a similar file for STF3A data.) For the example, 
we assume you are using seven Census variables, five from the first 
record and two from the second. (Each "logical" record of the STF1A 
data is split into two physical records of 4805 bytes each.  These
are referred to in the Census documentation as "segments".) 

Since some Census variables are character variables, which result
in completely blank spaces if missing, the output program specifies
column beginning points using the @1, @7, @12 (etc.) pointers, so
that the data 'line up' regardless of width or missing values.
An example of the output follows the program.
***********************************************************************/

 FILENAME RAW '~datastor/census/census1990/census90.stf1a.ca';
    DATA  CENSUS;
    INFILE RAW LRECL=4805;
    INPUT FILEID $ 1-8 STUSAB $ 9-10 SUMLEV 11-13 
        GEOCOMP $ 14-15 CHARITER 16-18 / 
        H13_1 2389-2397 H19 2767-2775;
 FILENAME SUBSET 'my.raw.census.subset';
    DATA _NULL_; SET CENSUS; FILE SUBSET;
    PUT @1 FILEID @10 STUSAB @15 SUMLEV @20 GEOCOMP 
        @25 CHARITER @30 H13_1 @40 H19;   RUN;

/*********************************************************************
(NOTE that even if you are not reading any variables from the second 
physical record or segment, you still must have the slash in the INPUT
statement so that SAS knows there are two physical records per observation.)

To use the above file for your own work, you need change only the INPUT
statement and the PUT statement.  For the INPUT statement, simply list
the variables and column specifications (with the $ if you're reading
an alphanumeric variable) from the Census code book for those variables
you want to read.   For the PUT statement, just list the same variable
names (but no columns or $, unless you need to have the values lined up
in columns) that you used in your INPUT statement.  

To run the program, type it into a UNIX file, then enter

  sas  <file.name>

SAS will create the raw data file 'my.raw.census.subset' (or whatever
you decide to call it).  The contents of this file, as generated by
the example above, will look like this (only the first ten observations 
are shown):

  STF1A    CA   040  00   0    457885    29008161
  STF1A    CA   040  40   0    764       45312
  STF1A    CA   040  42   0    0         0
  STF1A    CA   040  43   0    0         0
  STF1A    CA   050  00   0    17950     1242068
  STF1A    CA   060  00   0    890       68633
  STF1A    CA   070  00   0    890       68633
  STF1A    CA   080  00   0    8         3288
  STF1A    CA   091  00   0    5         1190
  STF1A    CA   091  00   0    0         813


You can then use this raw file as input to the stat package of
your choice.  
**************************************************************/