


Compute the constants for `tube-formula' based simultaneous confidence bands.
Usage: kap=kappa0(x,y,varargin)
Works for regression models only. Density estimation problems
should be converted to counts, and use poisson regression
'family','poisson'.
Essentially, this is a front-end to locfit, and so all optional
arguments to locfit (eg, smoothing parameters) can be provided.
To compute (or plot) the confidence bands, provide the output
of the kappa0() function as the 'kappa' argument to a
predict() or lfband() call.
Example:
load ethanol;
fit = locfit(E,NOx,'alpha',0.5)
kap = kappa0(E,NOx,'alpha',0.5) % give same arguments!
lfplot(fit)
lfband(fit,'kappa',kap) % plot the simultaneous bands
z = predict(fit,[0.6 0.7 0.8]','kappa',kap,'band','g')
z{3} % evaluate the bands.

0001 function kap=kappa0(x,y,varargin) 0002 % Compute the constants for `tube-formula' based simultaneous confidence bands. 0003 % 0004 % Usage: kap=kappa0(x,y,varargin) 0005 % 0006 % 0007 % Works for regression models only. Density estimation problems 0008 % should be converted to counts, and use poisson regression 0009 % 'family','poisson'. 0010 % 0011 % Essentially, this is a front-end to locfit, and so all optional 0012 % arguments to locfit (eg, smoothing parameters) can be provided. 0013 % 0014 % To compute (or plot) the confidence bands, provide the output 0015 % of the kappa0() function as the 'kappa' argument to a 0016 % predict() or lfband() call. 0017 % 0018 % 0019 % Example: 0020 % 0021 % load ethanol; 0022 % fit = locfit(E,NOx,'alpha',0.5) 0023 % kap = kappa0(E,NOx,'alpha',0.5) % give same arguments! 0024 % lfplot(fit) 0025 % lfband(fit,'kappa',kap) % plot the simultaneous bands 0026 % z = predict(fit,[0.6 0.7 0.8]','kappa',kap,'band','g') 0027 % z{3} % evaluate the bands. 0028 0029 fit = locfit(x,y,'what','kappa','ev','grid','mg',20,varargin{:}); 0030 z = fit{4}{6}; 0031 d = size(fit{1}{1},2); 0032 kap = z(1:(d+1)); 0033 0034 return;