|Date: ||Thu, 13 Oct 2011 21:38:54 -0400|
|Reply-To: ||R B <ryan.andrew.black@GMAIL.COM>|
|Sender: ||"SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>|
|From: ||R B <ryan.andrew.black@GMAIL.COM>|
|Subject: ||Re: Pilot study|
|Content-Type: ||text/plain; charset=ISO-8859-1|
You have over a 100,000 subjects and you're worried about power? I
doubt you're under-powered. Depending on what you're trying to do, you
might want to split your sample for validation purposes, but that's a
whole other topic.
On Thu, Oct 13, 2011 at 3:29 PM, William Shakespeare
> This will be retrospective in nature using data already collected. The
> questions I'm addressing all require logistic regression. I need to make
> sure I have enough cases to adequately power my study but coming up with
> the inputs for the sample size calculation is difficult. If this were
> prospective I'd pilot to obtain the inputs. Due to the retrospective
> nature of the present study the only way I could "pilot" it is to use the
> existing data. I don't know if that's appropriate.
> On Thu, 13 Oct 2011 10:27:15 -0700, Fareeza Khurshed <fkhurshed@GMAIL.COM>
>>Is it going to be a prospective study or a retrospective where you use
>>from the db?
>>On Thu, Oct 13, 2011 at 7:19 AM, William Shakespeare <
>>> I have a large database (>100k subjects) which I'm going to use to
>>> several research questions. Normally, with a prospective study I'd run
>>> pilot to obtain the necessary data for sample size calculation.
>>> that's not possible here. Would it be acceptable to draw a sample from
>>> database and derive pilot data from it as long as I exclude all those
>>> from subsequent analyses?