HOW TO PROCEED:
1) Copy the file src/ToMyOwn.c from the installation-directory
of nls2
and insert in it the required programs:
see comments included.
2) Open an S session and invoke `loadnls2'
with adding an argument called "tomyown",
equal to the pathname of this file.
(The programs may be split into several files: in that case,
the names of these files should be separated with spaces).
4) Invoke `nls2' with the
argument method set to "MYOWN",
and with an additional argument called num.ctx, equal to a
list-structure which contains the following components:
"nh": the number of sufficient statistics Z ("nh" in the `nls2 Reference Manual').
"effic": TRUE if the estimator is efficient, FALSE otherwise.
"W.type": a character string equal to "SYM" if the matrix W is symmetric,
"SYMBLOCK" if it is block symmetric and "NONSYM" in the other cases.
"fit.type": a character string that indicates how to calculate
the statistical criterion. Valid values are:
"LOGV": the statistical criterion is equal to -2log(likelihood),
"STOPCRIT": it is equal to the stopping criterion,
"NWSST": it is equal to the non-weighted sum of squares,
"VWSS": it is equal to the variance-weighted sum of squares,
"IVWSS": it is equal to sum of squares weighted by the intra-repetitions-variance,
"NWSSB": it is equal to the non-weighted sum of squares
in the context of the variance estimation, divided by n,
"SIGMA2": it is equal to square sigma,
"MYOWN": it is calculated by programs provided in "ToMyOwn.c"
(the default).
If there are several steps,
the elements of "method" that correspond to a step
where a
personal method is to be applied must be equal to "MYOWN",
and "num.ctx" is a vector of length equal to the number of
these elements.
The user's programs are provided in the file "ToMyOwn.c".See also the example nls2.ownmethod.
An S-session is opened with commands of the following pattern: loadnls2(tomyown="ToMyOwn.o") nls2.out <- nls2(data, model, stat.ctx, method = "MYOWN", num.ctx=list(nh=3,effic=T,W.type="SYM",fitting.crit="STOPCRIT"))
REFERENCE:
`nls2 Reference Manual'.
SEE ALSO:
`nls2',
the example nls2.ownmethod.