Replies to post #1656 on Dendreon Corp fka DNDNQ

[nerdseeksblonde]

25.9 is provenge overall median,
see my earlier post on the integral equations.
If you relate the median to the probability distributions,
you can get a simple expression in terms of KM curves but
it isn't always easy to solve.
There are 3 medians: the composite M, and the medians of
the two groups that make up the composite- M1 and M2 with
M1<=M2. The program finds M1 given a pdf, M, and M2.

It turns out that the result is very sensitive to the
fit to the KM curves, at least when you have two
nearby medians and one far away.
I'm not even sure the result survives if you make
the offset more accurate and subtract it from the
median :
// should approx the KM curve at time t months with median M.
mytype linP(mytype t, mytype M)
{
mytype off=4.5;
if (t<off) return 1;
t=t-off;
M=M-off; <<<<<<<<<<<<<<, This was missing before.
mytype x=1-t/M/2;
if (x<0) x=0;
return x;
}