Date: Thu, 30 Oct 2008 22:32:38 -0700
Reply-To: David L Cassell <davidlcassell@MSN.COM>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: David L Cassell <davidlcassell@MSN.COM>
Subject: Re: power analysis for joint modeling
Content-Type: text/plain; charset="iso-8859-1"
A power analysis in a situation like this can be phenomenally complicated, because the question becomes:
"Under what alternative hypotheses am I testing these ideas?"
There are a LOT of alternative hypotheses and alternative model specifications that you could use here. Do you know which ones are of interest in the power analysis you want to see?
(Bear in mind that traditional power analyses typically have simple models with normal errors and the only alternative of interest is changing the true mean of the normal. That simplifies things.)
David L. Cassell
3115 NW Norwood Pl.
Corvallis OR 97330
> Date: Mon, 27 Oct 2008 10:37:53 -0700
> From: ejshoptaw@YAHOO.COM
> Subject: power analysis for joint modeling
> To: SAS-L@LISTSERV.UGA.EDU
> Good afternoon,
> I was wondering if anyone could help me with a power
> analysis for a study using financial data. The data
> is severely skewed because many members do not occur
> any financial costs and then there are members that
> have extremely high costs. So a approach that I have
> seen is joint modeling.
> - Two arms, each having the same number of people
> - The joint modeling will have two sections:
> - The people without any financial cost will be
> modeled as a logit
> - The people will financial cost will be modeled
> using a gamma distribution.
> Is there any way to do a power analysis using joint
> modeling or is there a rule of thumb that is commonly
> Thanks in advance,
Want to read Hotmail messages in Outlook? The Wordsmiths show you how.