topic Re: Question about correspondence analysis in Statistical Procedures

Question about correspondence analysis

jhs2171 — Tue, 13 Dec 2016 19:03:37 GMT

I posted a question eariler today about different statistical methods to examine relationship between binary/nominal variables in my dataset. Someone suggested correspondence analysis and here's the SAS code I used.

data tab1;
input sleep_place$1-12 co_sleep$13-16 onback ontummy onside;
datalines;
SafePlace No 17 26 6
SafePlace Yes 1 2 1
UnsafePlace No 14 10 4
UnSafePlace Yes 44 27 20
;

data tab2;set tab1;
length newvar $ 28;
newvar=cat(sleep_place,co_sleep);run;

ods graphics on;
proc corresp data=tab2 plot;
var onback ontummy onside;
id newvar;run;
ods graphics off;

My question is: Does it make sense to have one combined ID variable? Also, is there anything I should add/delete/change?

Thanks!

Re: Question about correspondence analysis

Rick_SAS — Tue, 13 Dec 2016 19:08:05 GMT

That choice of ID variable doesn't seem to be helpful. Try combining them, like this:

data tab1;
length ID $16;
input sleep_place $1-12 co_sleep  $13-16 onback ontummy onside;
ID = catc(sleep_place,co_sleep);
datalines;
SafePlace   No  17 26 6
SafePlace   Yes 1 2 1
UnsafePlace No  14 10 4
UnSafePlace Yes 44 27 20
;

proc corresp data=tab1;
var onback ontummy onside;
id ID;
run;

Re: Question about correspondence analysis

jhs2171 — Tue, 13 Dec 2016 19:20:04 GMT

Hi Rick,

Thank you for your comment! After I posted my question, I realized it didn't really make sense to use only one of the variables as ID, so I combined the first two and made the change to my original posting! I just tried what you suggested, and it returned the same result so thank you for the confirmation!! You been super helpful today!