Use the SET statement to:

1

• Use the SET statement to
• create an exact copy of a SAS dataset
• modify an existing SAS dataset by creating new
  variables, subsetting (using a subsetting IF)
• concatenate datasets i.e., stack them one on top
  of the other. Most useful when the datasets
  contain the same variables but different
  observations
• interleave datasets i.e., keep sorted datasets
  sorted use a BY statement with the SET
  statement
• the general form is
• DATA new_data_set
• SET old_data_set
• other statements as needed

2

• make permanent SAS dataset from the data here
• time of train, of cars, total of riders
• then create new dataset with new variable
  average number of riders per car see page 171
  
• 1010 6 21
• 1215 10 56
• 1530 10 25
• 1130 8 34
• 1315 8 12
• 1045 6 13
• 2030 6 32
• 2315 6 12

3

• CONCATENATION OF TWO DATASETS - NOTE WHAT
  HAPPENS WHEN THEY HAVE DIFFERENT VARIABLES
• DATA southentrance
• INPUT Entrance PassNumber PartySize Age _at__at_
• DATALINES
• S 43 3 27 S 44 3 24 S 45 3 2
• PROC PRINT DATA southentrance TITLE 'South
  Entrance Data'
• DATA northentr ance
• INPUT Entrance PassNumber PartySize Age Lot _at__at_
• DATALINES
• N 21 5 41 1 N 87 4 33 3 N 65 2 67 1 N 66 2 7 1
• PROC PRINT DATA northentrance TITLE 'North
  Entrance Data'
• DATA both
• SET southentrance northentrance
• IF Age . THEN AmountPaid .
• ELSE IF Age
• ELSE IF Age
• ELSE AmountPaid 12

4

• INTERLEAVE datasets to keep an ordering in
  them
• The datasets to be INTERLEAVED must be SORTED
  by the interleaving variable
• DATA S SET southentrance
• PROC SORT DATAS BY PassNumber run
• PROC PRINT DATA S TITLE 'South Entrance Data'
• DATA N SET northentrance
• PROC SORT DATA N BY PassNumber
• PROC PRINT DATA N TITLE 'North Entrance
  Data'
• now put them together in order of passnumber
• DATA interleave
• SET N S BY PassNumber
• PROC PRINT DATA interleave
• TITLE 'Both Entrances, By Pass Number' RUN
  QUIT
• Now concatenate the two therapy datasets creating
  a new total number of patients variable
• first, without sorting then interleave them, BY
  month.

5
One-to-one match merge

• Use the MERGE statement with the BY statement to
  combine two datasets with a common so-called
  matching variable (the BY-variable) to uniquely
  identify each observation in the datasets...
• DATA new MERGE old1 old2
• BY matching_vars
• NOTE If the two datasets have variables with the
  same names (besides the BY variables) the
  variables from the second dataset will overwrite
  the values of the variables in the first dataset
  with the same names. Also note that all
  observations from both datasets are included in
  the new dataset whether they had a match or not...

6
Try to merge the demographic information about
the patients with the dataset on the visits they
made to the doctors office (ID, age, sex, date
of birth) A001 31 M 05/22/75 A002 43 M
06/15/63 A003 34 F 08/17/72 A004 . F
03/27/69 A005 54 F 02/24/52 A007 49 M
11/01/57 Next is the visits data (ID, visit,
sysBP, diasBP, Weight, date of visit)
7
A001 1 140 85 195 01/05/01 A001 2 138 90
198 10/13/01 A001 3 145 95 200 07/14/02 A002
1 121 75 168 04/14/01 A003 1 118 68 125
08/12/01 A003 2 112 65 123 11/15/01 A004 1
143 86 204 03/30/01 A005 1 132 76 174
02/27/01 A005 2 132 78 175 08/02/01 A005 3
134 78 176 12/04/01 A008 1 126 80 182
05/22/01 Whats the merging variable? Do you
need to sort by that variable before merging?
Can you merge them in descending order?? Try the
next one too - see section 6.4 on p. 176-177.
8
dm log 'clear' dm output 'clear' options
ls80 DATA sales INPUT CodeNum 1-4 PiecesSold
6-7 DATALINES C865 15 K086 9 A536 21 S163
34 K014 1 A206 12 B713 29 DATA
descriptions INPUT CodeNum 1-4 Name 6-14
Description 15-60 DATALINES A206 Mokka
Coffee buttercream in dark chocolate A536 Walnoot
Walnut halves in bed of dark chocolate B713
Frambozen Raspberry marzipan covered in milk
chocolate C865 Vanille Vanilla-flavored rolled
in ground hazelnuts K014 Kroon Milk chocolate
with a mint cream center K086 Koning Hazelnut
paste in dark chocolate M315 Pyramide White with
dark chocolate trimming S163 Orbais Chocolate
cream in dark chocolate
9

• PROC SORT DATA sales
• BY CodeNum
• PROC SORT DATA descriptions
• BY CodeNum
• Merge data sets by CodeNum
• DATA chocolates
• MERGE sales descriptions BY CodeNum
• PROC PRINT DATA chocolates
• TITLE "Today's Chocolate Sales" RUN quit
• NOTE the SORTing of the two datasets by the
  merging variable... all observations from both
  datasets are included in the merged dataset,
  whether they had a match or not.

10
One-to-many match merge

• Same MERGE statement is used as in one-to-one ...
  but the result will be different if you have many
  observations in one dataset to be matched with a
  single observation in the other...
• A good example of this is merging summary
  statistics with original data from which the
  statistics were computed. Go over sections
  6.5-6.6 to see how this is done... Also note the
  output (p. 181) from the PROC PRINT when both a
  BY and an ID statement are used together.
• Make the two shoes programs work (in section
  6.5 and 6.6).... use the code below and get the
  data from the book datasets and put them into
  files that the INFILE statements can read

11
dm log 'clear' dm output 'clear' options
ls80 DATA shoes INFILE '' INPUT Style
1-15 ExerciseType sales PROC SORT DATA
shoes BY ExerciseType DATA discount
INFILE '' INPUT ExerciseType Adjustment
Perform many-to-one match merge DATA prices
MERGE shoes discount BY ExerciseType NewPriceRO
UND(RegularPrice-(RegularPriceAdjustment),.01) P
ROC PRINT DATA prices TITLE 'Price List for
May' RUN quit
12

• Now consider an important merging application
  putting summary statistics back in the same
  dataset with the data used to compute the
  summaries. Note this is a many-to-one merge,
  since there are only a few statistics computed
  (one for each by-group value) and they are to be
  merged back with the many individual values used
  to compute those statistics
• As an example, compute the sum of the sales
  variable for each of the three classes of shoes
  in the previous example, and then merge that
  total back into the original dataset. After this
  you can compute the percentage of the total that
  each group represents, for example (get the
  sales data from page 180)

13

• PROC MEANS can be used to compute a grand total
  for a variable, but then it can't be merged back
  into the original dataset since you don't have a
  common BY variable in the two datasets Thus we
  must use a different technique (see section 6.7)
  it turns out that the SET statement with an IF
  - THEN will do the job
• DATA new if _n_1 then SET summary_data_set
• SET original_data_set
• The original dataset has many obs., while the
  summary dataset has only one. SAS reads this one
  obs. with the SET statement, but only for the
  first obs (i.e., when _n_ 1) this works
  because SET automatically RETAINS the
  observations from the first read - go over the
  example given in section 6.7 Combining a Grand
  Total with the Original Data

14

• For Monday In the padgett data, get the mean
  and standard deviation of plant height for the
  pizza hut" and shell island" marshes. Merge
  those statistics back into the original dataset
  and then calculate a z-score for each plant in
  the dataset. Recall that Z (X - mean)/(s.d.) ,
  where X is the original height. There are two
  means and two standard deviations, one for each
  of the two marshes pizza hut" and shell island
  and so you want to be sure that the correct mean
  and s.d. gets used with plants from the
  respective marshes