I've had to weight to weight some data that i'm working on so that both groups are balannced, my weightin efficiency is aroun 58% - i know that this isn't great but what level should i be looking for for my data to remain reliable?
2006-11-23 22:10:34 · 3 answers · asked by lucy1404 1 in Science & Mathematics ➔ Mathematics
It's not completely to do with the weighting efficiency itself. It depends how large your actual sample size is. If you only had 3 people in the sample then it wouldn't be very reliable regardless of the weighting; if you have 3 million people in the sample then it'd still probably be as reliable as you could hope for even with the heaviest weighting.
I think (I'm not sure of this, but fairly sure) that your effective sample size is your achieved sample size multiplied by the weighting efficiency, so multiply your sample size by 0.58
Eg. if you had 1,000 in your sample, then your results have an effective sample size of 580, so your answers would be as reliable as if you'd asked 580 people, so use this base size in significance tests rather than the full 1,000. ("reliability" in this case would rely on the % confidence you want to rely on, the actual percentages that you're comparing, and the effective sample sizes of the groups you're comparing).
2006-11-24 00:04:54 · answer #1 · answered by andyblacksheep 2 · 0⤊ 0⤋
If you are adding or subtracting the numbers, the absolute error is the sum of the absolute errors.
If you are multiplying or dividing, the relative error is the sum of the relative errors.
2006-11-25 15:53:35 · answer #2 · answered by Anonymous · 0⤊ 0⤋
yes
2006-11-24 06:18:25 · answer #3 · answered by dream theatre 7 · 0⤊ 0⤋