In a previous post , I mentioned about expected value and variance of different distributions.
Taking the same statistical concepts further, we now want to compute confidence intervals for our estimate.
- While thinking about Confidence Intervals, it is a good exercise to identify what distribution is representative of your estimate.
- The reason this is needed is because the confidence interval is dependent on standard deviation. As such, it would be necessary to know how you are computing your standard deviation.
- An alternative would be if we compute the variance from base principles.
- Here are a couple of very interesting post that explains the relationship between confidence intervals, statistical levels and P-values in a very simple way.
- CTR: Here are some interesting discussions around computing confidence intervals for a metric like CTR
- 2 videos here are helpful in understanding the basic concept