[R] gnls or nlme : how to obtain confidence intervals of fitted values
Pierre MONTPIED
montpied at nancy.inra.fr
Mon Oct 4 09:24:50 CEST 2004
Thanks Spencer but the intervals function gives confidence intervals of
the parameters of the model not the predicted values. In the Soybean
example it would be the CI of predicted weight for a given time, knowing
all the parameters (Asym, xmid, scal, variance function and residual)
and their distributions.
And what I need to calculate is precisely this CI.
My question is therefore is there an analytical way to calculate such
CI, whatever the model, or could I try some randomizing techniques such
as bootstrap or other ?
Thanks.
Spencer Graves wrote:
> Pinhiero and Bates (2000) Mixed-Effects Models in S and S-Plus
> (Springer) describe the use of "intervals" for that. In
> library(nlme), R 1.9.1 for Windows, I found documentation for
> intervals, intervals.gls, intervals.lme, intervals.lmList, and gnls.
> To an example in the documentation for gnls, I added intervals:
> > data(Soybean)
> > # variance increases with a power of the absolute fitted values
> > fm1 <- gnls(weight ~ SSlogis(Time, Asym, xmid, scal), Soybean,
> + weights = varPower())
> > summary(fm1)
> Generalized nonlinear least squares fit
> Model: weight ~ SSlogis(Time, Asym, xmid, scal)
> Data: Soybean
> AIC BIC logLik
> 983.7947 1003.900 -486.8974
>
> Variance function:
> Structure: Power of variance covariate
> Formula: ~fitted(.)
> Parameter estimates:
> power
> 0.8815437
>
> Coefficients:
> Value Std.Error t-value p-value
> Asym 17.35682 0.5226885 33.20682 0
> xmid 51.87232 0.5916820 87.66925 0
> scal 7.62053 0.1390958 54.78617 0
>
> Correlation:
> Asym xmid
> xmid 0.787 scal 0.485 0.842
>
> Standardized residuals:
> Min Q1 Med Q3 Max
> -2.309670367 -0.646844555 -0.004897024 0.498606088 4.986727281
>
> Residual standard error: 0.3662752
> Degrees of freedom: 412 total; 409 residual
> > intervals(fm1)
> Approximate 95% confidence intervals
>
> Coefficients:
> lower est. upper
> Asym 16.329331 17.356822 18.384313
> xmid 50.709200 51.872317 53.035434
> scal 7.347094 7.620525 7.893957
> attr(,"label")
> [1] "Coefficients:"
>
> Variance function:
> lower est. upper
> power 0.8369085 0.8815437 0.926179
> attr(,"label")
> [1] "Variance function:"
>
> Residual standard error:
> lower est. upper
> 0.3373374 0.3649392 0.3947995
>
> Is this what you want? hope this helps. spencer graves
>
> Pierre MONTPIED wrote:
>
>> Hi
>>
>> I use gnls to fit non linear models of the form y = alpha * x**beta
>> (alpha and beta being linear functions of a 2nd regressor z i.e.
>> alpha=a1+a2*z and beta=b1+b2*z) with variance function
>> varPower(fitted(.)) which sounds correct for the data set I use.
>>
>> My purpose is to use the fitted models for predictions with other
>> sets of regressors x, z than those used in fitting. I therefore need
>> to estimate y with (95%) confidence intervals.
>>
>> Does any body knows how to do this with R ?
>>
>> Thanks
>>
>> ______________________________________________
>> R-help at stat.math.ethz.ch mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-help
>> PLEASE do read the posting guide!
>> http://www.R-project.org/posting-guide.html
>
>
>
More information about the R-help
mailing list