[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

BGonline.org Forums

Confidence Interval Overlap vs Joint Standard Deviation (long)

Posted By: Timothy Chow
Date: Sunday, 1 November 2009, at 12:21 a.m.

In Response To: Confidence Interval Overlap vs Joint Standard Deviation (long) (neilkaz)

UBK "I also point out that the calculation of joint SD makes an independence assumption which I don't think is warranted."

It's true that the independence assumption isn't quite correct, and it's also true that the Gaussian approximation isn't quite correct. In my opinion there's no reason to fuss over these niceties, mostly because there isn't any computationally feasible way to calculate things exactly right anyway.

Neil asks how one should calculate jsd from rollouts ?

What you did was fine. Bob is just pointing out that it's an approximation.

As I said in my other reply, if you like to work with disjointness of confidence intervals, that's fine. It's not quite as theoretically satisfactory as JSD, but JSD isn't theoretically perfect either.

Messages In This Thread

 

Post Response

Your Name:
Your E-Mail Address:
Subject:
Message:

If necessary, enter your password below:

Password:

 

 

[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

BGonline.org Forums is maintained by Stick with WebBBS 5.12.