PROC SQL basics, tips and techniques and sample code programs

Proc SQL:


Power of SAS SQL:

• SQL looks at datasets differently from SAS
– SAS looks at a dataset one record at a time, using an implied loop that moves from the first record to the last
– SQL looks at all the records, as a single object
• Because of this difference SQL can easily do few things that are more difficult to do in SAS
• There are a number of built in functions in SQL that can be used in a select statement
• Because of how SQL handles a dataset, these functions work over the entire dataset
• Functions:

– Count: Counts Values
– Sum: Sums Values
– Max: Identifies the largest value
– Min: Identifies the smallest value
– Mean: Averages the values

Read more at www.cognigencorp.com/perspective/tipsNtricks.pub/1/PROC%20SQL%20Talk_12_.ppt –

SAS SQL

Introduction to Proc SQL

AN INTRODUCTION TO PROC SQL®

PROC SQL: When and How to Use It?

Proc SQL – A Primer for SAS Programmers

Understanding PROC SQL

Creating Macro Variables with Proc SQL

DATA Step vs. PROC SQL: What’s a neophyte to do?


Effective Use of SQL in SAS Programming

Variable List Short-Cuts in PROC SQL

Ten Best PROC SQL Tips and Techniques

Undocumented and Hard-to-Find PROC SQL C2 AE_Features

Variable List Short-Cuts in PROC SQL

Exploring the World of PROC SQL® Joins

Using PROC SQL to Create Ad Hoc Reports

CREATING MACRO VARIABLES VIA PROC SQL

PROC SQL for DATA Step Die-hards

Calculating Changes and Differences Using PROC SQL —With Clinical Data Examples

Validating Data Via PROC SQL

Length of Numeric variables GT 8 in SAS| StudySAS BLOG

Q&A: numeric variables length more than 8? We all know that the default length of the numeric variables in SAS is 8 and if suppose I want to store a number lets say (12345678910, which has a length 11 to numeric variable) to variable total, what should I do?


What if the numeric variable digits are more than 12 digits and i want to store them all without any E values?

ANS)

The default length of numeric variables in SAS is 8 and all the numbers that we see in the sas datasets are called as floating numbers(floating point binary) and not a regular sequence numbers form 1 to 10. When we are using SAS/Windows as our operating system and then the minimum length for any numeric variable should be 3(not 1 as we get confused all the time). So if a variable contas less than 3 digits means it is stored with less space.

The reason is, since a numeric variable will need a power and and the sign(+ or -), if SAS want to store a numberit defenitely needs a minimum of 3 bytes.

Depending upon the operating system we are using for SAS, the range for numeric variables can be 2 to 8 or 3 to 8.

Since the type of floating-point values is upto 16 decimal digits. we can store numbers upto 16(1234567891012234) total 16 digits for a numeric variable, but that can be possible if we use a format statement.

Ex:
data dsn;
a=1234567891012234;
format a best16.;run;

 If we open the output, we can see all the 16 digits were stored for the variable a exactly in the dataset dsn.

If the the value of numeric variable is upto 12 disgits we don't require to specify any formats, if it is more than 12 digits we have to mention specify the format statement. without it we will see error in the system log.

Options VALIDVARNAME=UPCASE

VALIDVARNAME= V7 UPCASE ANY

VALIDVARNAME= option is generally used in SAS whenever we want to control the SAS variable names in the dataset.

VALIDVARNAME= V7 UPCASE ANY

The default option will be VALIDVARNAME=V7 until we specify as UPPERCASE or ANY.

When we mention options VALIDVARNAME=V7, that means we are telling SAS to change the name of the Database column (etc EXCEL sheet column) to valid SAS name with certain rules keeping in mind.

Here are those rules that SAS needs to follow, when it changes the DBMS column name to valid SAS name.

Only 32 mixed case (lower or uppercase) characters are allowed in each variable.

Names should be starting with an underscore or an alphabet (either uppercase or lower case character).

Invalid characters in the DBMS column (ex. $) should be changed to underscores.

See the SAS Language Reference: Dictionary to get more details about the rules.

VALIDVARNAME=UPCASE
When we mention options VALIDVARNAME=UPPERCASE we are telling SAS to change the column name of the Database column to uppercase variables irrespective of type of variables in the DBMS column.


And whenever we want the same kind of characters in SAS dataset which are in the DBMS column (ex .(=) sign and the Asterisk(*) or the forward slash(\) we have to mention options

VALIDVARNAME=ANY
If we do, this will allows any characters which are in the DBMS column to be kept in the SAS dataset.

To understand the concept better here I am giving the example:

Example
The following example shows how the Pass-Through Facility works with
VALIDVARNAME=UPPERCASE.

options validvarname=uppercase;
proc sql;
connect to oracle as tables(user=USERID orapw=passward path=’INSTANCE’);
create table lab as
select lab_rslt, lab_test
from connection to oracle
(select "laboratory result$", "laboratory test$"
from DBMStable);
quit;

When we check the Output we observe that the variables in the DBMS column is changed to upper case as well as V7 (default option) converts those variables into UPPERCASE variables. Ex: " laboratory result$" becomes LAB_RSLT and " laboratory test$" becomes LAB_TEST.

How to merge data sets with a common variable?

Here is the simple way of merging the data sets with a common variable if the datasets has the same prefix name.

For example: col1-col10, dsn1-dsn 7 , or data1 to data10 with common variable of ID.

Considering we have 10 datsets and all of them having the same prefix data;

%macro mymerge (n);

data merged;
merge %do i = 1 % to &n;
data&i %end; ; /* this additional ';' is necessary, the first ';' is for the "%end", while the second ';' is for "Merge"*/;
by id;
run;
%mend;
%mymerge(10)

Merging the data sets using macro code

Merging the data sets with a common variable if the datasets has the same prefix name?

For example: col1-col10 dsn1-dsn 7 data1 to data6 with common variable of Usubjid.

here is the example, I have 7 datasets i need to merge and each of them having the common variable(usubjid) to merge, and all the datasets having the same prefix dsn(dsn1 to dsn7).

%macro allmerge (n);
data combine;
merge
%do i = 1 % to &n;
dsn&i
%end;
; /* this additional ';' is necessary, the first ';' is for
the "%end", while the second ';' is for "Merge" */
by usubjid;
run;
%mend;
%allmerge (7)

Resolving and Using &&var&i Macro Variables

Here are the 2 important questions always comes up in our minds,(& vs && vs &&& and single dot and double dots) when we are dealing with macros for the first time and here are the answers for them.

I did find a very good regarding the above topics in the one of the SAS forums and IAN WHITLOCK explained it very clear.

direct link :

or

when to use &,&&,and &&&,how do we distinguish?

&MACVAR references a macro variable. The rule is that the scanner reads from left to right. Two ampersands resolve to one and scanner continues. If the resulting expression has ampersands then that expression is scanned again. So &&x resolves to &x resolves to value of X

&&&x resolves to & followed by the value of X which then resolves to the value of the variable indicated.

If you are familiar with TSO CLISTS (even if you are not), they are very similar to SAS macro. SAS was originally based on PL1, so both SAS syntax and SAS macro are similar in some ways to

PL1 and PL1 macro.

what is the diff between Single dot and double dot(eg. &chech.> and &check..)

&CHECK and &CHECK. are the same thing. If the scanner finds a . that ends the macro variable reference, then the scanner eats the. and ends the reference. If there is no current macro variable then a . is a . So if &X is MYFMT then &X.. is MYFMT.

Test example: what is &&&X..5 when X has the value V and V has the value TEST?

&&&X..5 -> &V.5 -> TEST5

Test: Consider &&&...&X where the dots indicate there are a total of 15 ampersands preceding the X. Write a sequence of %LET statements followed by

%put &&&...&X ;

so that this causes text to be written on the log without any notes, warnings. or errors. One, three, and 15 are interesting sequences of amprsands. Find one number in between 3 and 15 and next oone after 15 that are interesting for the same reason. How many dots would be needed to make the letter S immeadiately follow the value of &&&...&X?

You may also need to check the SIGI paper to understand the multiple ampersands concept in detail:

Example: ....

%let dsn=study;
%let n=05;
%let dsn05=Client; %put &dsn&n; *Resolves to study05;

%put &dsn.&n; *Resolves to study05;

%put &dsn..&n; *Resolves to study.05;

*Multi Ampersands Concept; %put &&dsn&n;

* First Scan Resolves to &dsn05;
* Second Scan Resolves to Client;

%put &&&dsn&n;

* First Scan Resolves to &client05;
* Second Scan Resolves to &Client05; *with an error message in the log file....;

*Log file; 57 %put &&&dsn&n;

WARNING: Apparent symbolic reference STUDY05 not resolved.

&study05

/*Because.. the most common mistake is that.. */
/*We assume that macro variable resolution process proceeds from right to left...*/
/*Infact it isn't. */

The bottom line is ...

/**For some reason.... the &&&dsn&n is taken as && &dsn&n, which resolves to &client05.*/
/*This macro variable will not get resolved, because &client5. macro variable isn't there in the symbol table.; */

Macrovariables and its resolution:

Example:

%let one=two;
%let two=three;
%let three=Check;

%put &one;
%put &&one;
%put &&&one;
%put &&&&one;
%put &&&&&one;
%put &&&&&&one;
%put &&&&&&&one;
%put &&&&&&&&one;
%put &&&&&&&&&one;
%put &&&&&&&&&&one;

Answer:

%put &one; two
%put &&one; two
%put &&&one; three
%put &&&&one; two
%put &&&&&one; three
%put &&&&&&one; three
%put &&&&&&&one; Check
%put &&&&&&&&one; two
%put &&&&&&&&&one; three
%put &&&&&&&&&&one; three

Resolving and Using &&var&i Macro Variables

How can I count number of observations per subject in a data set?

We always have this question in mind, while we do the SAS programming and here is the simple answer for that, we just need to use SUM statement and the FIRST.variable in the SET statement and then the RETAIN statement to calculate the observations count per subject.

By doing some minor modification we can calculate observations count per subject per visit also. (Just include visit variable in the BY variable list in PROC sort and First. variable list in datastep with SET statement).


For example:

data dsn;
input patid implants;
datalines;
1 3
1 1
1 2
1 1
2 1
2 2
3 1
4 2
3 1
4 5
2 3
1 6
;
run;

proc sort data=dsn;
by patid;
run;

data dsn1;
set dsn;
by patid;
cnt+1;
if first.patid then cnt=1;
run;

proc sort data=dsn1;
by patid descending cnt;
run;

data dsn2;
set dsn1;
by patid;
retain totcnt;
if first.patid then totcnt=cnt;
output;
run;

proc print data=dsn2;
run;

Output: