[R] inverse prediction and Poisson regression

Vincent Philion vincent.philion at irda.qc.ca
Fri Jul 25 14:20:09 CEST 2003

Hello, and thanks for this.

>  From "www.r-project.org" -> search -> "R site search" -> "LD50", I found 
> "dose.p", described on p. 193, sec. 7.2, of Venables and Ripley (2002) Modern 
> Applied Statistics with S, 4th ed. (Springer).

I found the same, but this is for logistic regression I think, not Poisson. This is used to calculate the dose which causes 50% "death". I could do it this way by "pretending" my data is binomial.
i/e Y/Ymax = some value between 0 and 1. I could "pretend" my response data is "dead" or "alive", but I'm not sure this is "proper". but maybe it is? Any hints on this?

Y and Ymax follows a Poisson distribution. Can I just use Y/Ymax, run a logistic regression and go on with my life???

(LD50P100 <- dose.p(fitP100, p=14))
>              Dose         SE
> p = 14: -2.451018 0.04858572

Wow, you got the dose.p function to work with poisson regression? but under SPlus only? is it giving the right results?

... my 
> present insomniac state (3:20 AM in California).

The same for me!!!

I worked on this 2:00AM EDT
thanks and have a nice day, possibly starting with a good coffee!!!

> M. PHILION:  Does this provide sufficient information for you to now solve 
> your problem?

If this works under R and is giving the right solution, I owe you BIG TIME !

bye for now,

Vincent Philion, M.Sc. agr.
Institut de Recherche et de Développement en Agroenvironnement (IRDA)
3300 Sicotte, St-Hyacinthe
J2S 7B8

téléphone: 450-778-6522 poste 233
courriel: vincent.philion at irda.qc.ca
Site internet : www.irda.qc.ca

More information about the R-help mailing list