Re: [hkrug] sample size of paired t-test

[C.H.]

The effect size (d) is enormously big (d=-14, well, d = 1 is generally
considered a "large effect". Is the d=14 really the case?)
The pwr.t.test will break when d > 11 (and sig.level =0.05,
power=0.8), by then the estimated sample size will be <2 pairs. I
don't think you would like to have a study with only 1 pair. Then it
will be a case report, by the way.

On Tue, Jan 29, 2013 at 10:27 AM, Sara Fung <saras...@gmail.com> wrote:
> Hi all, I have trouble running sample size calculation for paired t-test,
> like that:
>
> pwr.t.test(d=-0.84/0.06, sig.level=0.05, power=0.8, type="paired",
> alternative="two.sided")
> Error in uniroot(function(n) eval(p.body) - power, c(2 + 1e-10, 1e+07)) :
> f() values at end points not of opposite sign
>
> what seems to be the problem? Please help, thanks. s
>
> --
> You received this message because you are subscribed to the Google Groups
> "Hong Kong R User group" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to hkrug+un...@googlegroups.com.
> To post to this group, send email to hk...@googlegroups.com.
> Visit this group at http://groups.google.com/group/hkrug?hl=en.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>

[Sara Fung]

Dear Hong, thanks for your reply.  Yes, it is the stupidity of me messing up the SE and SD, I realize it after I posted the help so I deleted it immediately, sorry for wasting your time! s