Profile Method for MS Objects

DESCRIPTION:

Returns a list of elements (data frames), one element for each profiled parameter.

USAGE:

profile.ms(fitted, which, npoints, zetamax, delta, trace=F, data) 

REQUIRED ARGUMENTS:

fitted
a previously fitted object inheriting from class "ms".

OPTIONAL ARGUMENTS:

which
the subset of the parameters to use for profiling; by default, all parameters are used.
npoints
the number of steps to take in profiling. The algorithm attempts to step npoints in each direction from the original fit, giving a total of 2*npoints + 1 new fits per parameter.
zetamax
the maximum absolute change in the objective function. Default 10.
delta
the vector of initial step sizes for each parameter. Defaults to a relative change of .5/npoints.
trace
should tracing be done during the re-fit. Defaults to FALSE, regardless of the trace control during the original fit.
data
optional data frame for fitting. Only needed if the data was not specified explicitly in the original fit.

VALUE:

a profile, as described under the generic function; that is, a list containing a data frame for each profiled parameter. The data frames have two variables, zeta and par.vals. The first is the square-root of the increase in the objective function from the global fit, signed by the change in the profiled parameter. The second is the matrix of corresponding profile-minimum parameter values, optimized over all the other parameters in the model for a fixed value of the profiled parameter.

EXAMPLES:

param(pingpong, "p") <- 0 
attach(pingpong, 1) 
D <- winner - loser 
p <- sum(winner > loser)/length(winner) 
alpha <- log(p/(1 - p))/mean(D) 
detach(1, save = "pingpong") 
lprob <- function(lp)log(1 + exp(lp)) - lp 
fit.alpha <- ms( ~ lprob(D * alpha), pingpong) 
profile(fit.alpha)