I am interested in obtaining the variance for measures of
association in 2 x 2 tables.
In particular, for a table with entries a, b, c, d,
V = (ad - bc) /sqrt((a+b)(a+c)(b+d)(c+d))
is a well-used measure (known as delta by geneticts looking at
linkage disequlibrium).
Kendall and Stuart (vol 2, 1974 edition) give a formula for var(V), but
it is not symmetric in b and c, which seems counter-intuitive.
I am unable to derive it myself or find another source of
the formula.
Any pointers would be much appreciated.
Cathryn Lewis
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|