Explores the behavior of the objective function near the solution.
This function is an S Version 3 generic (see Methods);
method functions can be written to handle specific
S Version 3 classes of data. Classes which already
have methods for this function include:
ms
,
nls.
USAGE:
profile(fitted, which, maxpts, ...)
REQUIRED ARGUMENTS:
fitted
the original fitted model object.
OPTIONAL ARGUMENTS:
which
which of the original parameters should be used in the profile.
By default, all parameters are included.
maxpts
the maximum number of points that should be used in profiling an
individual parameter
...
additional parameters, including a suggested change in the parameter
values, on a suitable scale for the kind of model.
For nonlinear least-squares,
delta.t is the suggested change on the
scale square-root of the change in sum-of-squares divided by the residual
sum of squares.
In this case, the interpretation of the change in terms of a t distribution
provides another argument,
alphamax, to limit the change in the
sum-of-squares to that significance level of the distribution.
VALUE:
a list, one element for each parameter varied.
The elements are data frames, with two variables.
One variable is the matrix of values for the parameters in the new fitted
models, the other is the corresponding values of the profile statistic
(a scaled version of the square-root difference between the objective
function and the original minimum).
REFERENCES:
Kaderman, E. and Bates, D. (1990).
Three-dimensional contour projections from objective functions,
Computer Science and Statistics: Proceedings of the 22nd Symposium on the Interface.
pp 382-386, Springer Verlag.