Hi again and thanks a lot for all your replies. You have provided me with a
lot of useful information and I more or less know how am going to deal with it.
I will use the Chebycheff's inequality to provide a lower bound for the
probability of error:
P(|Xs-X|<e)>= 1- (W (Xs)/e^2)
where :
X = population mean
Xs= sample mean
e= error
W(Xs)= (N-n/ N) V(Xs)
N= total population
n= sample size
This will give them some information. Most of those probabilities are >=
0.5 with e=1, which doesn't help much. But in that way I will be able to
show them how innacurate results are when there hasn't been a propper survey
planning beforehand. Suggesting to use stratified sampling the next time.
Thanks again for all your comments
Luis
|