I’ve now made that spreadsheet viewable by anyone with the link. If you want to use it you should save a copy you can edit.
It doesn’t estimate regression as such; rather it’s for keeping track of the inside-view uncertainty about model parameters, in a way which is quick and easy but roughly correct. This gets you to something like a total inside view of the error inherent in the model; if you also have a prior over effectiveness you could then use it to regress.
I talk more about the details in this post. The spreadsheet was made mostly for my own use when I want to make quick estimates with error-bars on log-normal distributions; I’ve tried to make it slighly user-friendly but I’m afraid it’s not terribly so.
You might be interested in this spreadsheet we designed to help estimate regression to the mean:
https://docs.google.com/spreadsheet/ccc?key=0AtqNQZ7WlsnLdDdfZ1BseWgxNHhyMnR6cllxbmhFUlE&usp=drive_web#gid=0
Owen can explain better than me how it works.
I’ve now made that spreadsheet viewable by anyone with the link. If you want to use it you should save a copy you can edit.
It doesn’t estimate regression as such; rather it’s for keeping track of the inside-view uncertainty about model parameters, in a way which is quick and easy but roughly correct. This gets you to something like a total inside view of the error inherent in the model; if you also have a prior over effectiveness you could then use it to regress.
I talk more about the details in this post. The spreadsheet was made mostly for my own use when I want to make quick estimates with error-bars on log-normal distributions; I’ve tried to make it slighly user-friendly but I’m afraid it’s not terribly so.