[R] ed50

[Rubén Roa]

> -----Mensaje original-----
> De: r-help-bounces at r-project.org 
> [mailto:r-help-bounces at r-project.org] En nombre de Dipa Hari
> Enviado el: lunes, 12 de julio de 2010 22:19
> Para: r-help at r-project.org
> Asunto: [R] ed50
> 
> I am using semiparametric Model
>  library(mgcv)
> sm1=gam(y~x1+s(x2),family=binomial, f)
> 
> How should I  find out standard error for ed50 for the above 
> model ED50 =( -sm1$coef[1]-f(x2)) / sm1$coef [2]
>  
> f(x2) is estimated value for non parametric term.
>  
> Thanks

Two ways, 

1) Re-parameterise the model so that ed50 is an explicit parameter in the model, or
2) Taylor series (aka Delta method) using the standard errors of coef[1], coef[2] and their correlation.

HTH
____________________________________________________________________________________ 

Dr. Rubén Roa-Ureta
AZTI - Tecnalia / Marine Research Unit
Txatxarramendi Ugartea z/g
48395 Sukarrieta (Bizkaia)
SPAIN