# RE: st: reverse prediction - confidence interval for x at given y in nonlinear model

 From "Daniel Waxman" To statalist@hsphsun2.harvard.edu Subject RE: st: reverse prediction - confidence interval for x at given y in nonlinear model Date Fri, 26 Oct 2007 11:50:00 -0700

```Nick,

Unfortunately I am away from a computer with Stata installed and have been tapping this out on a cell phone!

if dose=0 then the logistic equation reduces to

b[zdose] + b[cons]

If dose=1, it reduces to just

b[cons].

Agreed.  Units are irrelevant.

I'll tell you what:  When I get home on 11/1, i'll start a new thread with a more convincing example (unless somebody wants to agree with me before then)...

Dan

-----Original Message-----

From:  "Nick Cox" <n.j.cox@durham.ac.uk>
Subj:  RE: st: reverse prediction - confidence interval for x at given y in nonlinear model
Date:  Fri Oct 26, 2007 11:23 am
Size:  2K
To:  <statalist@hsphsun2.harvard.edu>

We have been at cross purposes.

Your log(0) as 0 is, I see, only a trick that keeps
the zeros in the frame _so long as you also have a dummy
for 0_.

My point is that it does not help in problems in which there is no dummy
for
0.

Suppose your doses are 0 0.1 0.2 0.3 0.4 0.5
in whatever units are being used. Neither Stata nor the mathematics

Then the recipe cond(dose == 0, 0, log(dose)) is not even monotonic
in dose. 0 is treated as if were larger than any other positive
value in the data!

Also if any dose is 1, this recipe treats 0 and 1 as
equivalent.

Nick
n.j.cox@durham.ac.uk

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
```