Macros: Code cutters or code generators

1
Macros Code cutters or code generators

• By
• Prashant Mittal
• pmittal_at_usm.maine.edu

2
Introduction

• Macros Code cutters or code generators?
• Why I learnt Macros

3
Macro Structure

• macro macro_name
• mend macro_name
• macro_name
• run
• Regular SAS data steps.
• Procedure statements.
• Macro Statements.

4
SAS OOP or not

• Function z f ( x, y)
• Java, C, C
• ? Have built in functions.
• ? Users may define their own functions.
• ? Reuse these functions by storing them in a
  library.

• SAS
• ?Has built in procedures.
• (Proc Means, Proc Freq. Proc Reg etc.)
• ? Users may define their own functions in the
  form of macros.
• ? Reuse these functions by storing them in a
  library.

5
Macros written as Functions

• macro xyz (a , b ,c )
• data any
• set any1
• newvar1 a
• newvar3 b
• newvar3 c
• run
• mend xyz
• xyz (a var1, b var2, c var3)
• xyz (a var4, b var5, c var6)
• xyz (a var7, b var8, c var9)

6
Macro variables and their scope

• Tools that enable to dynamically modify the text
  in a SAS program though symbolic substitution are
  Macro Variables.
• Macro variables are default characters.
• They can be temporarily converted to numeric for
  calculations.
• User defined macro variables Automatic macro
  variables.
• Global macro variables Local Macro Variables

7
Macro variables and their scope

• To get description of macro variables that exist
  at a particular time
• Put _all_ _automatic_ _global_ _local_
  _user_
• To write instructions in SAS log, the following
  code could be used
• ? Put str ( string)
• The above code will print string in SAS log.
• ? Put Report for sysday, sysdate generates
  the following in SAS log
• Report for Thursday 02APR04

8
Creating new Macro Variables

• let macro_variable This is it
• put macro_variable
• The above two lines will produce This is it
• in SAS log.

• let a Example
• data new
• do i 1 to 10
• var ranuni(0)
• newvar "a"
• output
• end
• run
• proc print data new
• run
• Obs i var newvar
• 1 1
  0.83883 Example
• 2 2
  0.82309 Example
• 3 3
  0.20670 Example
• 4 4
  0.48918 Example

9
Creating Macro Variables

• data team2
• input position 12. player 12.
• call symput ('POS' left (_n_), position)
• cards
• batter Ann
• pitcher Tom
• frstbase Bill
• run
• data team2
• set team2
• new "pos1"
• run
• proc print data team2
• Title Newly created macro variables are pos1,
  pos2, pos3"
• run

• Newly created macro variables are batter ,
  pitcher , frstbase
• 1903 Thursday, April 1, 2004
• Obs position player new
• 1 batter Ann batter
• 2 pitcher Tom batter
• 3 frstbase Bill batter

10
Description of the dataset used

• Medicaid claims file.
• Originally had more than 2.5 million
  observations.
• Randomly chose 1000 observations.

11
Writing repetitive code using Macros

• Proc means n min max mean sum median std P50 data
  any
• Var charg1 Paid1
• Class elig1
• Run
• Proc means n min max mean sum median std P50 data
  any
• Var charg2 Paid2
• Class elig2
• Run
• Proc means n min max mean sum median std P50 data
  any
• Var charg3 Paid3
• Class elig3
• Run
• .
• .
• How wonderful would it be to write a do loop to
  handle these repetitive procedures

12
Writing repetitive code using Macros

• Do loop is not restricted to the data step.
• macro easy
• do i 1 to 15
• Proc means n mean sum median std data any
• Var chargi Paidi Indirect referencing
• Class eligi
• run
• end
• mend
• easy
• run

13
Writing repetitive code using Macros Example 1

• Proc means mean median std data any
• Var charg1 Paid1
• Class elig1
• Run
• Proc means mean median std data any
• Var charg2 Paid2
• Class elig2
• Run
• Proc means mean median std data any
• Var charg3 Paid3
• Class elig3
• Run
• .
• .
• .
• .

• macro easy
• do i 1 to 7
• Proc means mean median std data any
• Var chargi Paidi Indirect referencing
• Class eligi
• run
• end
• mend
• easy
• run

14
Writing repetitive code using Macros Example 2

• proc sort data any
• by elig1
• run
• proc freq data mainecare95_02
• table chg1paid1
• by elig1
• title' Cross tabulation between chg1 and paid1'
• run
• proc sort data any
• by elig2
• run
• proc freq data mainecare95_02
• table chg2paid2
• by elig2
• title' Cross tabulation between chg2 and paid2'
• run
• .
• .

• The previous repetitive code of more than 150
  lines could be written as follows
• macro easy
• do i 1 to 15
• proc sort data any
• by eligI
• proc freq data any
• table chgipaidi
• by eligi
• title' Cross tabulation between chgi and
  paidi'
• run
• end
• mend
• easy
• run

15
Repetitive case where variable names are not
sequential

• Suppose you want to have descriptive
  statistics for the variables in line1 by
  variables in the line 2.
• Line 1 A B C D E
• Line 2 V W X Y Z

16
Writing repetitive code using Macros Example 3,
Case where variables are not sequential.

• Proc means mean median std data any
• Var A
• Class V
• Run
• Proc means mean median std data any
• Var B
• Class W
• Run
• Proc means mean median std data any
• Var C
• Class X
• Run
• Proc means mean median std data any
• Var D
• Class Y
• Run

• macro easy (var1 , var2 )
• Proc means mean sum median std data any
• Var var1
• Class var2
• Run
• mend
• easy(var1 A, Var2 V)
• easy(var1 B, Var2 W)
• easy(var1 C, Var2 X)
• easy(var1 D, Var2 Y)
• easy(var1 E, Var2 Z)
• run

17
Output chi-sq of multiple cross-tabs directly to
a dataset.

• Consider an example where the experimenter is
  looking to get only the chi-square values of
  cross tabulations of a variable say A, with 12
  other variables (B, C, D, E, F, G, H, I, J, K, L,
  M).
• Regular code will be written the following way
• proc freq data any
• table A( B C D E F G H I J K L M) /CHISQ EXACT
• run

18
Output chi-sq of multiple cross-tabs directly to
a dataset.

• data final
• set _null_
• proc freq data any noprint
• table AB/ chisq
• output PCHI out a
• data a
• Set a
• vars "A-B"
• data final
• set final a
• run
• proc freq data any noprint
• table AC/ chisq
• output PCHI out a

• (continued..)
• proc freq data any noprint
• table AD/ chisq
• output PCHI out a
• data a
• Set a
• vars "A-D"
• data final
• set final a
• run
• proc freq data any noprint
• table AC/ chisq
• output PCHI out a
• data a

19
Output chi-sq of multiple cross-tabs directly to
a dataset.

• data final
• set _null_
• macro chisq(var1 ,var2 )
• proc freq data any noprint
• table var1var2/chisq
• output PCHI out a
• data final
• set final a
• vars "var1-var2"
• run
• mend
• chisq(var1 A,var2 B)chisq(var1 A,var2
  C) chisq(var1 A,var2 D)
• chisq(var1 A,var2 E)chisq(var1 A,var2
  F)chisq(var1 A,var2 G)
• run

20
Time trend analysis

• ? Study unpaid amount by Medicaid with time.
• ? Unpaid amount Charges Payments.
• ? To study if there is a time trend in unpaid
  amounts.
• ? Procedure Break down the whole dataset into
  groups of 50 (say), sort them by year and study
  the descriptive statistics of unpaid amount.