Lower bound and upper bound

Avatar

Avatar

Kushsoni

Avatar

Kushsoni

Kushsoni

09-01-2018

Target Question.png

Hi All, I am not able to understand how to calculate upper and lower bound in above question. Thanks in advance.

Thanks & Regards, Anshul Singhal

Replies

Avatar

Avatar

Amelia_Waliany

Employee

Avatar

Amelia_Waliany

Employee

Amelia_Waliany
Employee

24-01-2018

Hi there and thanks for your question! You may find Confidence Level and Confidence Interval in the Adobe Target Product Documentation helpful. Also, be sure to check out the Adobe Target Sample Size Calculator and this article on Estimating Lift in Revenue.

Hope this helps!

Avatar

Avatar

Andrey_Osadchuk

MVP

Avatar

Andrey_Osadchuk

MVP

Andrey_Osadchuk
MVP

27-02-2019

Can anyone explain how to calculate the lower and upper bound for the question attached by Kushsoni?

Avatar

Avatar

clewisAdobe

Avatar

clewisAdobe

clewisAdobe

28-02-2019

Kushoni and Andrey,

With the details of the question provided above, you don't have enough information to calculate the standard deviation and therefore, cannot get an exact margin of error and the confidence interval. But you can eliminate some of the answers with a solid understanding of the pieces at play here.

For Experience B, there are two options for the Lower Bound of the Confidence Interval. Both cannot be correct, so we will decide which answer is correct between these two. With RPV as the KPI, we're looking at the +1.2% RPV LIFT with only 61% confidence (1 minus the p-value of 0.39). That 1.2% is the calculated mean, such that the Lower Bound must fall below the mean by definition. Therefore we can rule out answer A with +2%.

Similarly, for Experience C, there are two options for the Upper Bound. The higher confidence interval will be wider, as it means the interval needs to capture a larger portion of the population. We also know that Experience C is currently sitting at a strong 94% confidence level, so the -4.3% RPV LIFT is relatively trustworthy. With both of those considerations, Answer D with upper bound of the 90% confidence interval at +3% is beyond the more reasonable -1% for the 95% confidence Interval. So we eliminate Answer D.


And so, to answer the question around the calculating the lower and upper bound for the confidence interval, you'll need the mean of the data (RPV in this case) and the standard deviation of the data (average of the squared differences from the mean, for each order). From there, using the desired confidence level and sample size, you can calculate the margin of error, which give you the mean +/- values for the upper and lower bound.

I hope that helps.

Avatar

Avatar

Andrey_Osadchuk

MVP

Avatar

Andrey_Osadchuk

MVP

Andrey_Osadchuk
MVP

01-03-2019

clewisAdobe​, thank you very much for the detailed explanation. Now it's all clear