I would like to received some suggestions about this subject;
I need a confidence interval for a difference of proportion (p1- p2) that
have been calculated from a weight formula such as p1=sumh(wh*p1h) and
p2=sumh(wh*p2h) where wh is the weight of the h
stractum (h=1,...,H) and pih (i=1,2;h=1,...,H) is the proportion of elements
in the sample i and stratctum h that fall in a class C.
The interval that i achieved, for a significance level of alfa is,
( (p1-p2) (+ or -) z(1-alfa/2) sqrt(var(p1-p2)) )
where, var(p1-p2)=sumh( (wh)^2 * ph * qh * (1/n1h +1/n2h), and ph=(n1h * p1h
+ n2h * p2h)/(n1h+n2h)
in the formula nih (i=1,2;h=1,...,H) is the number of elements from i
sample and h stractum.
What i ask is;
is this formula rigth? and if is not, do you know some references about
this problem.
Thanks in advance, best regards
___________________________________
Baltazar Nunes
Statistician
Obeservatório Nacional de Saúde
Instituto Nacional de Saúde Dr. Ricardo Jorge
Av. Padre Cruz, 1649-016 Lisboa - Portugal
e-mail - [log in to unmask]
URL www.insarj.pt
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|