Margins of error

#1
Hi all, please help - I understand that if I have a binary proportion (i.e. % of Yes & % of No), than I can caluculate margin of error based on STDV and n.
But how about if I have % of A, % of B, % of C etc? Should I treat them as % of A and NOT A %, then % of B and NOT B? Can I calculate margin of error when I have multiple proportions? Thank you beforehand!
 
#5
There's something quite unsettling about all this. The errors aren't independent because they must sum to zero.
%C = 100% - %A - %B. So you might imagine that moeC = sqrt(moeA^2+moeB^2). But the same applies to %B and %A...?
Very strange. But the suggestions above will probably give reasonable results.
 

obh

Active Member
#6
It must support reasonable result as if you only have the data of A and not A, this is the correct model. You can only say that the may be better model for this case since you have more data.