[R] svydesign syntax and deviance!

[Thomas Lumley]

On Sat, 31 Jul 2010, aline uwimana wrote:

> Thank you Thomas,
>  if i understand what you say below i need to use glm in order to have the deviance and consider it for the model with svyglm and choose the best model. I
> use  glm and svyglm for two models but don't have deviance for the svyglm. As i need to choose with the two models it is correct to use the deviance from
> glm model for the svyglm.

No. It isn't correct to use the deviance at all.  The deviance is based on the likelihood for independent sampling, and that is not what you have.

    -thomas

Thomas Lumley
Professor of Biostatistics
University of Washington, Seattle