intervals.nnr {assist}  R Documentation 
Approximate posterior standard deviations are calculated for the spline estimate of
nonparametric functions from a nnr
object, based on which approximate Bayesian
confidence intervals may be constructed.
intervals.nnr(object,level=0.95, newdata=NULL, terms=<see below>, pstd=TRUE,level=0.95, ...)
nnr.obj 
an object inheriting from class nnr , representing a
nonlinear nonparametric regression model fit.

newdata 
a data frame on which the fitted spline estimates are to be evaluated.
Only those predictors, referred in func of nnr fitting, have to
be present. The variable names of the data frame should correspond
to the function(s)' arguments appearing in the opion func= of nnr.
Default is NULL, where predictions are made at the same values
used to fit the object.

level 
set as 0.95, unused currently 
terms 
an optional named list of vectors or matrices containing 0's and 1's collecting one or several combinations of the components of spline estimates in the fitted snr object. The length and names of the list shall match those of the unknown functions appearing in the 'snr' fit object. For the case of a single function, a vector of 0's and 1's can also be accepted. A value "1" at a particular position means that the component at that position is collected. Default is a vector of 1's, representing the overall fits of all unknown functions. 
pstd 
an optional logic value. If TRUE (the default), the posterior standard deviations are calculated. Orelse, only the predictions are calculated. Computation required for posterior standard deviations could be intensive. 
level 
a numeric value set as 0.95. 
... 
other arguments, currently unused. 
The standard deviation returned is based on approximate Bayesian confidence intervals as formulated in Ke and Wang (2002).
an object of class bCI
is returned, which is a list of length 2.
Its first element is a matrix which contains predictions for
combinations specified by terms
, and second element is a matrix which contains
corresponding posterior standard deviations.
Chunlei Ke chunlei_ke@pstat.ucsb.edu and Yuedong Wang yuedong@pstat.ucsb.edu
Ke, C. and Wang, Y. (2002). Nonlinear Nonparametric Regression Models. Submitted.
## fit a generalized varying coefficient models data(Arosa) Arosa$csmonth < (Arosa$month0.5)/12 Arosa$csyear < (Arosa$year1)/45 ozone.fit < nnr(thick~f1(csyear)+exp(f2(csyear))*f3(csmonth), func=list(f1(x)~list(~I(x.5),cubic(x)), f2(x)~list(~I(x.5)1,cubic(x)), f3(x)~list(~sin(2*pi*x)+cos(2*pi*x)1,lspline(x,type="sine0"))), data=Arosa, spar="m", start=list(f1=mean(thick),f2=0,f3=sin(csmonth)), control=list(backfit=1)) x < seq(0,1,len=50) u < seq(0,1,len=50) ## calculate Bayesian confidence limits for all components of all functions p.ozone.fit < intervals(ozone.fit, newdata=list(csyear=x,csmonth=u), terms=list(f1=matrix(c(1,1,1,1,1,0,0,0,1),nrow=3,byrow=TRUE), f2=matrix(c(1,1,1,0,0,1),nrow=3,byrow=TRUE), f3=matrix(c(1,1,1,1,1,0,0,0,1),nrow=3,byrow=TRUE))) ## Not run: plot(p.ozone.fit, x.val=x)