[R] mgcv: beta coefficient and 95%CI

[David Winsemius]

On Feb 22, 2011, at 11:34 PM, clc wrote:

>
> Hi i am doing an environmental research
>
> The equation is as follow:
>
> gam(y1 ~ x1 + s(x2) + s(x3) + s(x4), family = gaussian, fit = true)
>
> I would like to obtain the beta coefficient and 95CI of x4 (or  
> s(x4)), what
> should I do?

You should re-think your expectations. There is no "beta coefficient".  
The whole point of smooth terms in gam models is that the role of the  
beta coefficient in ordinary models is taken over by a function that  
varies across the range of x4. If yu want to see its form, you need to  
specify the other terms and plot the results of predict(gam(.)) with  
varying x4.

>
> Thanks,
> Lung
> -- 
> View this message in context: http://r.789695.n4.nabble.com/mgcv-beta-coefficient-and-95-CI-tp3320491p3320491.html

==

David Winsemius, MD
West Hartford, CT