02-27-2014 03:53 PM
Has anyone ever noticed that there is an obvious error in the documentation for 9.22 on the page with the title, "Positional and Nonpositional Syntax for Coefficients in Linear Functions?"
Actually, I think there are two errors on the page, but I am less certain about the second one.
The documentation asserts that if factor A has three levels and factor B has two levels, then these two procedure calls are equivalent:
class a b;
model y= a b a*b;
estimate 'B at A2' b 1 -1 a*b 0 0 1 -1;
class b a;
model y=a b a*b;
estimate 'B at A2' b 1 -1 a*b 0 1 0 0 -1;
This cannot be correct since there are 4 numbers after a*b in the first example but 5 in the second. I suspect that one of the 0's needs to be removed. Confirmation?
The second suspicious thing on the page is that the "nonpositional" syntax changes depending on the position of a and b on the class statement. In what sense would such a syntax be nonpositional?
I think that in both instances, the correct nonpositional syntax is
estimate 'B at A2' b 1 -1 a*b [1 2 1] [-1 2 2];
02-27-2014 04:33 PM
Well, I just tested the code. Should have done that first. It does work. So I guess my questions are:
1) Why in the positional syntax are 5 values needed if B is listed first and just 4 if A is listed first?
2) Why does the nonpositional syntax depend on position?
02-27-2014 04:53 PM
I figured out the first question. With the positional syntax, there should be 6 coefficients, but SAS assumes that if you name less than 6 then however many at the end are missing, they are all put =0. Would be really nice if the documentation of this was improved.
0 0 1 -1
is the same as
0 0 1 -1 0 0
0 1 0 0 -1
is the same as
0 1 0 0 -1 0