Difference in proportion for large sample size


I have 2 proportions from 2 years:
P1(2001): 11000/12500 = 88.0%
P2 (2002): 11500/12800 = 89.8%

I am interested in knowing if there is a statistical increase in % from 2001 to 2002. So, H0 = P1-P2>0; H1 = P1-P2 < 0

I had applied z-test to this problem and it gave a significant result. Is this true? :confused:


TS Contributor
Yes, at such large sample sizes even minute differences will be significant. The question is, is this practically relevant?

Thanks for the reply. Will think about it. =)

Is there another method for comparing proportions for large sample size?

I have another set of %
Pre: (sample proportion) 6000/10000 = 60%
Post: (population proportion) 70000/100000 = 70%

Would it be correct to do a one way z-test using post 0.7 as expected value?


TS Contributor
sure, but you will get the same problem - you need domain knowledge to decode if the difference is important or not, as opposed to significant