|
Home > Statistics Every Writer Should Know > The Stats Board > Discusssion
Statistical Confidence We use a statistical difference test tool to draw conclusions from our online advertising tests. We enter in the data. For example, if we are testing banners' click rate (clicks/impressions) for 2 different banners, and we enter in the data, say 100 imps and 10 clicks for banner 1 and 105 imps and 11 clicks for banner 2. Then we choose a confidence level (50-95%) to see if the winner would be the winner all the time (based on the confidence level). Do you have any insight on what this confidence level exactly means? Thanks.
READERS RESPOND: Re: Statistical Confidence always be 10%. The rate will vary. What you want to know, is how much can you expect it to vary within the population of all banner 1 ratios. The Confidence Level will return an interval about which you can expect to see the real population rate for banner 1. From your example, it looks like you are comparing banner 1 to banner 2 and looking at the differences between the ratios. Let's say the difference between the two ratios is 5%. You will use this information to determine an interval about which you can expect the differences between the two to fall. If this interval contains 0 then you could conclude that there's no difference between banner 1 and banner 2. Generally, CI used are set to 95%. Let's say the interval calculates to -5% to +5%. What you would be saying is that you are 95% confident that the true proportional difference between banner 1 and banner 2 lies between -5% to +5%. Since 0 is not included in this interval, you can assume that there's no difference between the two banner rates.
Your $5 contribution helps cover part the $500 annual cost of keeping this site online.
|
|||||||||
|