topic Re: categorical variable in SAS Programming

categorical variable

Phamhhm — Thu, 15 Nov 2018 20:31:56 GMT

I have problem for coding character var in categorical var.

DATA Bla;

SET Bla;

IF area = "Chicago" THEN area = 1;

ELSE IF area = "New York City" THEN area = 2;

ELSE IF area = "DC/Maryland/Virginia" THEN area = 3;

ELSE area = . ;

RUN;

PROC FREQ DATA = Bla;

TABLE area;

RUN;

I don't get the categories in my table, jus a row with :

area Frequency Percent Cumulative Frequency Cumulative Percent

. 50000 100.00 50000 100.00

Thanks

Re: categorical variable

Reeza — Thu, 15 Nov 2018 20:38:36 GMT

You cannot change a variable type like that. And you destroyed your data so you need to start over and recreate the bla file first.

You cannot recode a variable from character (area) to numeric (area). Change the variable name for one and it will be fine.

Re: categorical variable

ballardw — Thu, 15 Nov 2018 20:49:27 GMT

@Phamhhm wrote:

Hi

I have problem for coding character var in categorical var.

DATA Bla;

SET Bla;

IF            area = "Chicago"                         THEN area = 1;

ELSE IF area = "New York City"               THEN area = 2;

ELSE IF area = "DC/Maryland/Virginia"    THEN area = 3;

ELSE                                                                     area = . ;

RUN;

PROC FREQ DATA = Bla;

TABLE area;

RUN;

I don't get the categories in my table, jus a row with :

area          Frequency          Percent         Cumulative Frequency          Cumulative Percent

.                50000                 100.00                 50000                              100.00

Thanks

You should look at your log from the data step. I bet it has message about character value was converted to numeric some where.

I think that your actual data in BLA before this step had a numeric variable named area. In that case when you did

if area='Chicago' then

SAS attempted to change the text "Chicago" to a numeric value to compare the actual value of the variable area. Since all of your comparisons shown would fail to convert to numeric the result of each comparison was "false" and the only assignment possible was the final assignment to missing.

Perhaps you had a different variable intended to compare such as CITY or similar?

But as @Reeza says the original data set BLA has been overwritten and may not be useable for your purpose.

Especially when recoding an existing variable the use of the

Data bla;

set bla;

is extremely dangerous as logic errors, such as in this example, have the potential of destroying you data.

Best is to create a new data set.

Repeated use of the structure also makes it extremely difficult to debug where something may have gone wrong as you overwrite the complete set at each use.

Re: categorical variable

Phamhhm — Thu, 15 Nov 2018 20:53:04 GMT

I have a copy of the data.

I did change the name of the variable, but still doesn't work.

narea = 1;

narea =2;

.....

But for my other variable, i have not problem by coding :

DATA Bla;

SET Bla;

IF x_type = "AAA" THEN x_type=1;

ELSE IF x_type = "BBB" THEN x_type=2;

ELSE x_type= .;

Re: categorical variable

Reeza — Thu, 15 Nov 2018 20:53:52 GMT

Check your log. There's a note in there regarding type conversion I bet. I suspect your area variable is already numeric with a format applied, check it via PROC CONTENTS.

Re: categorical variable

Phamhhm — Thu, 15 Nov 2018 20:56:00 GMT

I also did change the DATA set name and it doesn't work.

DATA Blabla;

SET Bla;

Re: categorical variable

Phamhhm — Thu, 15 Nov 2018 21:00:06 GMT

I start over, delete the data set, import the data set, give the name of the variable narea , and it does not work.

Re: categorical variable

Reeza — Thu, 15 Nov 2018 21:02:09 GMT

Post your full new code and the log then.
'does not work' does not provide any information that we can help you with. You need to provide explicit details.

Re: categorical variable

Phamhhm — Thu, 15 Nov 2018 21:07:08 GMT

Original data = Bla

var = area : Char

Data = Blabla

new var = narea : Num

Re: categorical variable

Shmuel — Thu, 15 Nov 2018 21:31:40 GMT

Please check your input by running

proc freq data=Bla; table area; run;

Are there the areas you are looking for? Do they match upcase/locase letters exactly ?

and as said before:

- change output dataset name to differ from input

- define the category variable name different from current variables in dataset

Re: categorical variable

Reeza — Thu, 15 Nov 2018 21:35:08 GMT

I'm not sure what you're trying to say. The variables appear to have the correct type, but if it's not working and you don't show your code, I can't help beyond what I've already stated. Good Luck.

Re: categorical variable

Phamhhm — Fri, 16 Nov 2018 00:41:46 GMT

1). SAS doesn't care about upper case or lower case.

2). Given a new name to a dataset is to prevent losing the original dataset.

3). Given a new name to a variable after transformation/manipulation is to prevent losing the old variable.

All the 3 little details are not the source of my problem, because I have try and you could try it to.

Anyways, here are the code , as you can see...

Thanks for the help.... People are not working all the time, people get out and go eat.

***********************************************************************************************************

DATA PHAM.Blabla;
SET PHAM.Bla;
IF div_type = "BTH" THEN div_type = 1;
ELSE IF div_type = "LDD" THEN div_type = 2;
ELSE IF div_type = "LTD" THEN div_type = 3;
ELSE div_type = .;
RUN;

***************************************************************************************************************

NOTE: Numeric values have been converted to character values at the places given by:
      (Line):(Column).
      93:42   94:42   95:42   96:42
NOTE: There were 50000 observations read from the data set PHAM.BLA.
NOTE: The data set PHAM.BLABLA has 50000 observations and 2 variables.
NOTE: DATA statement used (Total process time):
      real time           0.01 seconds
      cpu time            0.01 seconds

***************************************************************************************************************

PROC FREQ DATA = PHAM.Blabla;
TABLE div_type;
RUN;

***************************************************************************************************************

Version:1.0 StartHTML:000000271 EndHTML:000037427 StartFragment:000035746 EndFragment:000037389 StartSelection:000035746 EndSelection:000037377 SourceURL:file:///C:/Users/Minh%20Pham/AppData/Local/Temp/SAS%20Temporary%20Files/_TD2756_DESKTOP-MSHHIQ0_/sashtml.htmSAS Output

The SAS System

The FREQ Procedure

40721	81.44	40721	81.44
963	1.93	41684	83.37
7855	15.71	49539	99.08
461	0.92	50000	100.00

***************************************************************************************************************************************

DATA PHAM.Blabla;
SET PHAM.Bla;
IF        area = "Atlantic South"       THEN narea = 1;
ELSE IF   area = "California North"     THEN narea = 2;
ELSE IF   area = "Central/South Texas" THEN narea = 3;
ELSE IF   area = "Chicago"              THEN narea = 4;
ELSE IF   area = "Dallas"               THEN narea = 5;
ELSE IF   area = "DC/Maryland/Virginia" THEN narea = 6;
ELSE IF   area = "Great Lakes"          THEN narea = 7;
ELSE IF   area = "Houston"              THEN narea = 8;
ELSE IF   area = "Los Angeles"          THEN narea = 9;
ELSE IF   area = "Midwest"              THEN narea = 10;
ELSE IF   area = "New England"          THEN narea = 11;
ELSE IF   area = "New York City"        THEN narea = 12;
ELSE IF   area = "North Florida"        THEN narea = 13;
ELSE IF   area = "Northwest/Rocky Mountain" THEN narea = 14;
ELSE IF   area = "Ohio"                     THEN narea = 15;
ELSE IF   area = "Philadelphia"             THEN narea = 16;
ELSE IF   area = "South Florida"            THEN narea = 17;
ELSE IF   area = "Southwest"                THEN narea = 18;
ELSE IF   area = "Tennessee"                THEN narea = 19;
ELSE                                             narea = .;
RUN;

******************************************************************************************************************

NOTE: There were 50000 observations read from the data set PHAM.BLA.
NOTE: The data set PHAM.BLABLA has 50000 observations and 3 variables.
NOTE: DATA statement used (Total process time):
real time 0.03 seconds
cpu time 0.03 seconds

********************************************************************************************************************

PROC FREQ DATA = PHAM.Blabla;
TABLE narea;
RUN;

*********************************************************************************************************************

Version:1.0 StartHTML:000000271 EndHTML:000036897 StartFragment:000035746 EndFragment:000036859 StartSelection:000035746 EndSelection:000036821 SourceURL:file:///C:/Users/Minh%20Pham/AppData/Local/Temp/SAS%20Temporary%20Files/_TD2756_DESKTOP-MSHHIQ0_/sashtml.htmSAS Output

The SAS System

The FREQ Procedure

***********************************************************************************************************************************

PROC CONTENTS DATA = PHAM.Bla;
RUN;

********************************************************************************************************************************

Version:1.0 StartHTML:000000271 EndHTML:000041087 StartFragment:000035746 EndFragment:000041049 StartSelection:000035746 EndSelection:000041049 SourceURL:file:///C:/Users/Minh%20Pham/AppData/Local/Temp/SAS%20Temporary%20Files/_TD2756_DESKTOP-MSHHIQ0_/sashtml.htmSAS Output

The SAS System

The CONTENTS Procedure

PHAM.BLA	50000
DATA	2
V9	0
2018-11-15 19:15:41	33
2018-11-15 19:15:41	0
	NO
	NO

WINDOWS_64
wlatin1 Western (Windows)

65536
26
1
1977
1940
0
YES
C:\Program Files\Oxford\HEC\Techniques Exploitation Données\Travail2\bla.sas7bdat
9.0401M5
X64_10HOME
DESKTOP-MSHHIQ0\Minh Pham
2MB
1769472

area	Char	30	$F30.	$F30.	Area
div_type	Char	3	$F3.	$3.	Division Type Code

************************************************************************************************************************************

PROC CONTENTS DATA = PHAM.Blabla;
RUN;

************************************************************************************************************************************

Version:1.0 StartHTML:000000271 EndHTML:000041088 StartFragment:000035746 EndFragment:000041050 StartSelection:000035746 EndSelection:000041050 SourceURL:file:///C:/Users/Minh%20Pham/AppData/Local/Temp/SAS%20Temporary%20Files/_TD2756_DESKTOP-MSHHIQ0_/sashtml.htmSAS Output

The CONTENTS Procedure

PHAM.BLABLA	50000
DATA	3
V9	0
2018-11-15 19:25:27	48
2018-11-15 19:25:27	0
	NO
	NO

WINDOWS_64
wlatin1 Western (Windows)

65536
37
1
1361
1333
0
YES
C:\Program Files\Oxford\HEC\Techniques Exploitation Données\Travail2\blabla.sas7bdat
9.0401M5
X64_10HOME
DESKTOP-MSHHIQ0\Minh Pham
2MB
2490368

area	Char	30	$F30.	$F30.	Area
div_type	Char	3	$F3.	$3.	Division Type Code
narea	Num	8

Re: categorical variable

Shmuel — Fri, 16 Nov 2018 02:16:34 GMT

@Phamhhm wrote

1).  SAS doesn't care about upper case or lower case.

that's right dealing with variable names.

that's wrong dealing with variable value.

Re: categorical variable

Phamhhm — Fri, 16 Nov 2018 02:45:16 GMT

The dataset is available, the code and the log output are also posted as requested .

Does anyone know how to figure out the solution to my problem that can help a beginner in SAS....

Re: categorical variable

Phamhhm — Fri, 16 Nov 2018 03:16:17 GMT

Solution is found, problem is solved.

Thank you for all your advices.

Re: categorical variable

Shmuel — Fri, 16 Nov 2018 05:40:28 GMT

@Phamhhm, it will be nice if you post how did you solve the issue.

Re: categorical variable

Phamhhm — Fri, 16 Nov 2018 14:30:20 GMT

Just add ¨Area¨

DATA PHAM.Blabla;

SET PHAM.Bla;

IF area = "ATLANTIC SOUTH AREA" THEN area = 1;

ELSE IF area = "CALIFORNIA NORTH AREA" THEN area = 2;

ELSE IF area = "CENTRAL/SOUTH TEXAS AREA" THEN area = 3;

ELSE IF area = "CHICAGO AREA" THEN area = 4;

ELSE IF area = "DALLAS AREA" THEN area = 5;

ELSE IF area = "DC/MARYLAND/VIRGINIA AREA" THEN area = 6;

ELSE IF area = "GREAT LAKES AREA" THEN area = 7;

ELSE IF area = "HOUSTON AREA" THEN area = 8;

ELSE IF area = "LOS ANGELES AREA" THEN area = 9;

ELSE IF area = "MIDWEST AREA" THEN area = 10;

ELSE IF area = "NEW ENGLAND AREA" THEN area = 11;

ELSE IF area = "NEW YORK CITY AREA" THEN area = 12;

ELSE IF area = "NORTH FLORIDA AREA" THEN area = 13;

ELSE IF area = "NORTHWEST/ROCKY MOUNTAIN AREA" THEN area = 14;

ELSE IF area = "OHIO AREA" THEN area = 15;

ELSE IF area = "PHILADELPHIA AREA" THEN area = 16;

ELSE IF area = "SOUTH FLORIDA AREA" THEN area = 17;

ELSE IF area = "SOUTHWEST AREA" THEN area = 18;

ELSE IF area = "TENNESSEE AREA" THEN area = 19;

ELSE area = .;

RUN;

Re: categorical variable

Reeza — Fri, 16 Nov 2018 15:45:23 GMT

That will leave notes in your log that will make it difficult to find a true error.

Re: categorical variable

ballardw — Fri, 16 Nov 2018 17:40:47 GMT

@Phamhhm wrote:

Just add ¨Area¨

DATA PHAM.Blabla;

SET PHAM.Bla;

IF area = "ATLANTIC SOUTH AREA" THEN area = 1;

ELSE IF area = "CALIFORNIA NORTH AREA" THEN area = 2;

ELSE IF area = "CENTRAL/SOUTH TEXAS AREA" THEN area = 3;

ELSE IF area = "CHICAGO AREA" THEN area = 4;

ELSE IF area = "DALLAS AREA" THEN area = 5;

ELSE IF area = "DC/MARYLAND/VIRGINIA AREA" THEN area = 6;

ELSE IF area = "GREAT LAKES AREA" THEN area = 7;

ELSE IF area = "HOUSTON AREA" THEN area = 8;

ELSE IF area = "LOS ANGELES AREA" THEN area = 9;

ELSE IF area = "MIDWEST AREA" THEN area = 10;

ELSE IF area = "NEW ENGLAND AREA" THEN area = 11;

ELSE IF area = "NEW YORK CITY AREA" THEN area = 12;

ELSE IF area = "NORTH FLORIDA AREA" THEN area = 13;

ELSE IF area = "NORTHWEST/ROCKY MOUNTAIN AREA" THEN area = 14;

ELSE IF area = "OHIO AREA" THEN area = 15;

ELSE IF area = "PHILADELPHIA AREA" THEN area = 16;

ELSE IF area = "SOUTH FLORIDA AREA" THEN area = 17;

ELSE IF area = "SOUTHWEST AREA" THEN area = 18;

ELSE IF area = "TENNESSEE AREA" THEN area = 19;

ELSE area = .;

RUN;

So the problem was you wrote code comparing values of a variable that did not exist in your data if this works.

And now you will have notes about conversion to numeric

10   data example;
11      area = "ATLANTIC SOUTH AREA";
12      IF area = "ATLANTIC SOUTH AREA" THEN area = 1;
13   run;

NOTE: Numeric values have been converted to character values at the places given by:
      (Line):(Column).
      12:48

If the whole purpose of this exercise is to display those digits instead of the name text then a format would be my preferred solution as I could switch between the full text and the code value as desired:

data example;
   infile datalines dsd truncover;
   informat Area $30.;
   input area $;
datalines;
"ATLANTIC SOUTH AREA"
"CALIFORNIA NORTH AREA"
"CENTRAL/SOUTH TEXAS AREA" 
"CHICAGO AREA"
"DALLAS AREA"
"DC/MARYLAND/VIRGINIA AREA"
"GREAT LAKES AREA"
"HOUSTON AREA"
"LOS ANGELES AREA"
"MIDWEST AREA"
"NEW ENGLAND AREA"
"NEW YORK CITY AREA"
"NORTH FLORIDA AREA"
"NORTHWEST/ROCKY MOUNTAIN AREA"
"OHIO AREA"
"PHILADELPHIA AREA"
"SOUTH FLORIDA AREA"
"SOUTHWEST AREA"
"TENNESSEE AREA"
"Someplaceelse"
 ;
run;
proc format library=work;
value $area
"ATLANTIC SOUTH AREA" = '1'
"CALIFORNIA NORTH AREA"  = '2'
"CENTRAL/SOUTH TEXAS AREA"  = '3'
"CHICAGO AREA"  = '4'
"DALLAS AREA"  = '5'
"DC/MARYLAND/VIRGINIA AREA"  = '6'
"GREAT LAKES AREA"  = '7'
"HOUSTON AREA"  = '8'
"LOS ANGELES AREA"  = '9'
"MIDWEST AREA"  = '10'
"NEW ENGLAND AREA"  = '11'
"NEW YORK CITY AREA"  = '12'
"NORTH FLORIDA AREA"  = '13'
"NORTHWEST/ROCKY MOUNTAIN AREA"  = '14'
"OHIO AREA"  = '15'
"PHILADELPHIA AREA"  = '16'
"SOUTH FLORIDA AREA"  = '17'
"SOUTHWEST AREA"  = '18'
"TENNESSEE AREA"  = '19'
other = ' ' /* place what ever text you want in the quotes*/
;
run;

Proc print data=example;
   var area;
   format area $area.;
run;

Formats become more powerful if you have the value and the codes in a data set as you can make a format from the data set.

So instead of adding lots more If/then/else statements to your data code you just change the definition of the format. Also note that the Proc format code more compact than If/then/else.