YRBS DATA

[dfamu1989]

1229 proc corr data = yrbs2017;

1230 var age raceeth sex grade cigarette vaping;

ERROR: Variable RACEETH in list does not match type prescribed for this list.

1231 run;

Getting this error message when using
YRBS Data trying to code for Race/Ethnicity

[PGStats]

Variable RACEETH must be character. You will need to convert it to a number, otherwise it can't be involved in correlation calculations.

[dfamu1989]

race=" ";

if raceeth = "Am Indian/ Alaska Native" and raceeth ="Asian" and raceeth

="Native Hawaiian/other Pl" and raceeth ="Multiple-Non-Hispanic"

and raceeth= " Multiple - Hispanic" then race = "1";

if Raceeth = "White" then race = "2";

if Raceeth= "Black" then race= "3";

/* 1= other, 2=White, 3=black*/

(This is how I Have it coded)

[Shmuel]

use variable race instead raceet  in proc corr.

[dfamu1989]

When I replace it race I still got a error message.

603 proc corr data = yrbs2017;

604 var age race sex grade cigarette vaping;

ERROR: Variable race in list does not match type prescribed for this list.

605 run;

[dfamu1989]

How do you convert a character to a number

[dfamu1989]

data yrbs2017;

set DATAOUT.YRBS2017;

Sex=.;

if Q2=1 then sex=1;

if Q2=2 then sex=0;

/* 1=female, 2=male*/

race=" ";

if raceeth = "Am Indian/ Alaska Native" and raceeth ="Asian" and raceeth

="Native Hawaiian/other Pl" and raceeth ="Multiple-Non-Hispanic"

and raceeth= " Multiple - Hispanic" then race = "1";

if Raceeth = "White" then race = "2";

if Raceeth= "Black" then race= "3";

/* 1= other, 2=White, 3=black*/

age=.;

if Q1=1 then age=1;

if Q1=2 then age=2;

if Q1=3 then age=3;

if Q1=4 then age=4;

if Q1=5 then age=5;

if Q1=6 then age=6;

if Q1=7 then age=7;

grade=.;

if Q3=1 then grade=1;

if Q3=2 then grade=2;

if Q3=3 then grade=3;

if Q3=4 then grade=4;

cigarette=.;

if Q30=1 then cigarette=0;

if Q30=2 then cigarette=1;

/* 0=cigarette use, 1=no cigarette use*/

vaping=.;

if Q34=1 then vaping=0;

if Q34=2 then vaping=1;run;

/* 0=vaping, 1=no vaping*/

proc freq data=yrbs2017;

tables age raceeth sex grade cigarette vaping;

run;

proc freq data=yrbs2017;

tables vaping* age/ chisq;

run;

proc freq data=yrbs2017;

tables vaping* raceeth/ chisq;

run;

proc freq data=yrbs2017;

tables vaping* cigarette/ chisq;

run;

　

proc freq data=yrbs2017;

tables cigarette* vaping/ chisq;

run;

proc corr data = yrbs2017;

var age race sex grade cigarette vaping;

run;

[unison]

dfamu,

Taking an excerpt from your code...

Change the resulting coding from character to numeric i.e. change "1", "2", "3" to 1, 2, 3

*...;
if raceeth = "Am Indian/ Alaska Native" and raceeth ="Asian" and raceeth
="Native Hawaiian/other Pl" and raceeth ="Multiple-Non-Hispanic"
and raceeth= " Multiple - Hispanic" then race = 1;
if Raceeth = "White" then race = 2;
if Raceeth= "Black" then race= 3;
*...;

-unison

[mkeintz]

It's a straightforward programming task to convert a character variable race to a numeric variable.  But it is not quite as simple an analysis task.  In your case, you have 3 levels, not 2, which will not provide the information you want to get in proc corr.  Should you convert white/black/multiple-Hispanic to 0/1/2 or to  1/2/0, or …… or to 2/1/0 (6 possible combinations)?  Each would give a different value in the correlation matrix.  You would get three different positive values and their 3 negatives. The problem is how to interpret a correlation coefficient intended for linear variables (of which binary is a subset here) to a 3-category variable.

I presume you could make two dummy variables  (white/black  and white/multiple,  or black/while and black/multiple …).  Then you would have two directly interpretable correlation coefficients.   However, I leave it to more authoritative respondents to better explain my observation.  

[ballardw]

---

@dfamu1989 wrote:

/* 1=female, 2=male*/

race=" ";

if raceeth = "Am Indian/ Alaska Native" and raceeth ="Asian" and raceeth

="Native Hawaiian/other Pl" and raceeth ="Multiple-Non-Hispanic"

and raceeth= " Multiple - Hispanic" then race = "1";

if Raceeth = "White" then race = "2";

if Raceeth= "Black" then race= "3";

First to have a variable to use in Proc Corr it must be numeric. The statement race=" "; means that you just created a CHARACTER variable. So get rid of:race=" ";

Second you explicitly assign character values "1" "2" and "3", which again are not what you want.

And then third the logic in

if raceeth = "Am Indian/ Alaska Native" and raceeth ="Asian" and raceeth
="Native Hawaiian/other Pl" and raceeth ="Multiple-Non-Hispanic"
and raceeth= " Multiple - Hispanic" then race = 1;

Is NEVER going to be true. One person from your data cannot have any multiple values of Raceth. So instead of AND you would want OR

if raceeth = "Am Indian/ Alaska Native" 
or raceeth ="Asian" 
or raceeth ="Native Hawaiian/other Pl" 
or raceeth ="Multiple-Non-Hispanic" 
or raceeth= " Multiple - Hispanic" then race = 1; 

SAS has a comparison operator called IN to compare a single variable to a list of values to use instead of a stack of OR statements:

if raceeth  in( "Am Indian/ Alaska Native" "Asian" "Native Hawaiian/other Pl" 
                "Multiple-Non-Hispanic" " Multiple - Hispanic" )then race = 1; 

I am not sure but I believe that you may also have a typo in "Native Hawaiian/other Pl" I suspect that instead of a lower case L that you should have an upper case I at the end:  "Native Hawaiian/other PI"

[dfamu1989]

IT WORKED!!!!! THANKS SO MUCHHHHH

[ballardw]

---

@dfamu1989 wrote:

IT WORKED!!!!! THANKS SO MUCHHHHH

---

Nice.

Now for some possibly less than good news: You may want to talk with someone about the structure of the YRBS data sets and weighting. The YRBS typically is a complex sample design and if you use Proc Corr with the weights you are not correctly applying the sample design information and can get misleading results from Proc Corr..