On 3 Dec 2002 12:47:11 -0800, [EMAIL PROTECTED] (nothanks) wrote:
> Dear sci.stat*ers,
>
> I'm reading Collett's book on survival analsyis and cannot figure
> out the derivation for the # of individuals at risk in j-th time
> interval
> [t_j, t_{j+1}].
>
> This # (called effective sample size in SAS documentaion) is
>
> n_j + c_j/2,
>
> where n_j is number alive at j-th interval and c_j is number of
> censored survival times.
>
> Can anyone offer an explanation?
The effective sample is the number who went through
the period, plus half the dropouts. (The ones who 'died'
usually are credited with only 1/2 of that period, too.)
The assumption is that the dropouts (censored) are
dropping out, on the average, halfway through the period;
and another assumption is that the period is relatively
small -- so that the risk, in every sense, is about the same
in the first half <of the period> as in its second half.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
