Date: Wed, 13 Apr 2011 02:23:27 -0400
Reply-To: Søren Lassen <s.lassen@POST.TELE.DK>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Søren Lassen <s.lassen@POST.TELE.DK>
Subject: Re: OT: Jackknife estimates
Content-Type: text/plain; charset=ISO-8859-1
I do not know that much about jackknifing (except when diving), but
it seems to me that if your "point estimates" are mean values,
you may have made an error. As far as I can discern, the mean is not
affected by jackknifing (it should not be, as the numeric mean itself is
unbiased), what you can use the jackknife method for in terms of
means is to find outliers, as they may stand out more clearly when
you display the pseudovalues than when you display the actual values.
If you display an example of your code, maybe someone here can help you
look for errors.
On Mon, 11 Apr 2011 12:48:59 -0600, Reeza K <fkhurshed@HOTMAIL.COM> wrote:
>I'm trying to estimate some rates using a jackknife method but the rates
are really small, ie 165/5000. Using the leave one out method and
pseudovalues I end up getting jack knife estimates that are twice my actual
>Does anyone know if this is as a result of a small rate and if there are
any work arounds, I need to calculate confidence intervals. Since I really
only need the variance estimate does it matter if the jack knife estimate
isn't close to my point estimate, though that seems counterintuitive.
>Or alternatively if I'm doing something wrong....
>For the above example
>165/5000 the actual rate is 0.031933 but the jack knife rate is 0.064927.