[R] Prediction and confidence intervals from predict.drc

[David Winsemius]

On Sep 6, 2010, at 7:54 PM, Brant Inman wrote:

> R-helpers,
>
> I am using the package "drc" to fit a 4 parameter logistic model.   
> When I
> use the predict function to get prediction on a new dataset, I am not
> getting the requested confidence or prediction intervals.  Any idea  
> what
> is going on?  Here is code to reproduce the problem:
>
> ---
>
> library(drc)
>
> # Fit model to existing dataset in package
> spinach.model <- drm(SLOPE~DOSE, data = spinach, fct = LL.4())

(Comparing to the predict example code.) You did not specify a CURVE  
argument. Not sure what effect that would have.
>
> #Generate new fake dataset
> newdt <- data.frame(matrix(c(seq(0, 150, 0.1), rep(NA, 1501)), ncol=2,
> byrow=F))
> colnames(newdt) <- c('DOSE', 'SLOPE')

You did not include a CURVE variable. But this provokes nary a  
complaint. I wondered if the estimates may be an unlabeled mixture  
from the 5 CURVEs
>
> #Use predict function to get prediction and confidence intervals
> pred <- predict(spinach.model, interval='prediction', newdata=newdt)

You did not specify CURVE. The example uses one:
predict(spinach.model1, data.frame(dose=2, CURVE=c("1", "2", "3")),  
interval = "prediction")
With your object:

 > head(predict(spinach.model, data.frame(dose=2, CURVE=c("1")),
+ interval = "prediction"))
Prediction      Lower      Upper
  0.3500492 -0.2790351  0.9791336

With the original example:
 > predict(spinach.model1, data.frame(dose=2, CURVE=c("1", "2", "3")),
+ interval = "confidence")
      Prediction     Lower     Upper
[1,]  0.9048476 0.8552178 0.9544775
[2,]  0.4208307 0.3626741 0.4789873
[3,]  0.5581673 0.4971838 0.6191509
 >
With your object
 > predict(spinach.model, data.frame(dose=2, CURVE=c("1")),
+ interval = "confidence")
Prediction      Lower      Upper
  0.3500492  0.2673464  0.4327521

> conf <- predict(spinach.model, interval='confidence', newdata=newdt)
> head(pred); head(conf)
>
> ---
>
> Examining the output shows the point estimates but not the intervals.
> Would like the intervals.
>
>
> I am using the most recent versions of R and drc on Windows XP.
>
> Thanks,
>
> Brant
> 	[[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.