@bazingarollcall wrote: Thanks for your reply, ballardw. My sample size is actually 550274 and the race/ethnicity variables total to 551202, so some observations are presumably being counted more than once. How do I go about checking this?
I think that you need to write out some very explicit numbers stating where it comes from and how you adt them to get 551202. None of the data you showed in the first post does that.
If you are totaling all of the 1's across 4 variables then you have misunderstood what your code is doing. There is nothing in your code that made any of the assignments of the variables whtnh, blknh, hisp and else exclusive. Nothing. If you expect the 1 values to total to 550274 then your assignment code is wrong.
Use the Proc Freq code I showed before only add all 4 of your dependent variables in the tables statement:
tables nrace*hispanic*whtnh* blknh* hisp * else / list missing;
You will see that you have some values with 1 in multiple columns.
Note that not using the proper values can generate issues. When you use Nrace=2 in one place and Nrace not in ('0' '1' '2') then conversions of values become a possible issue. So, is NRACE numeric or Character?
The following data step shows what happens when coding with each combination of Nrace and Hispanic as provided assuming NO records are missing either of the variables.
data toofew toomany;
do nrace=0 to 5,8;
do hispanic= 1,2,9;
if (nrace=1) and (hispanic=2) then whtnh=1;
else whtnh=0;
if (nrace=2) and (hispanic=2) then blknh=1;
else blknh=0;
if nrace not in (1,2,0) and (hispanic=1) then hisp=1;
else hisp=0;
if nrace not in (1, 2, 0, 8) then else=1;
else else=0;
if sum(whtnh,blknh,hisp,else)>1 then output toomany;
if sum(whtnh,blknh,hisp,else)=0 then output toofew;
end;
end;
run;
Note that the data set TOOMANY will have three records that show that HISP and ELSE are both set to one. That might be part of your issue.
I would suggest that you look at the TOOFEW data set as well. That shows 9 records where the values of all the variables are 0. Which considering that some of then have Hispanic=1 may well not be what you expect.
I really do not understand why your race = MR (multiracial), white or black is excluded from HISP. Typically if you have a "non-Hispanic White" category then Hispanic White is counted in Hispanic, same with black or other races.
... View more