Posted by Ueli (22.214.171.124) on October 07, 1999 at 09:11:45:
In Reply to: first proposal posted by Riccardo Pentenero on October 07, 1999 at 03:37:03:
I see some flaws in your arguing:
Assuming there is no correlation between being a Chead and having access to the net, the figures in your examplemean:
60% of all the CHeads in the younger group have access to the net
30% of all the CHeads in the older group have access to the net
Now, I don't know which age group is attracted more to the xxxx sites and which is more likely to seek info abouttheir disease, so let's assume that the same fraction of CHeads with net access of both groups have found our site
If 100 people of the younger group answer your question and 20% report success with O2, then the conclusion to draw is: 20% of ALL CHeads in the younger group have success, regardless if they have access to the net or not, if they have found this site or not, if they answered your poll or not. At least, that is what statisticians do (and they would kindly, for not to confuse the reader, suppress the error margin, being 10% for a sample of 100; but they would have asked 99 or 101 people because this would give a few digits after the decimal point, suggesting a high precision).
In the same way 70.00000% of ALL CHeads in the older age group would have success.
The only valid conclusion to draw from the figures of your example is: O2 is much more effective for the older than for the younger.
If you would like to have a general average, you cannot weight the two groups by the number of responses, nor the fraction of net access, but you have to use the fraction of CHeads in the general population in each age group (a number that at the best could be an educated guestimate).
Your reasoning reminds me of those politicians, who after losing the election say: If the weather hadn't be so fine on election day, more people would have cast their vote (to us and not to the other party). ;^)
Post a Followup