Table of contents:
-
Syntax and structure of the input file
-
How to describe the:
-
Example
Each piece of information is introduced by an identifier.
Most of them can be omitted
(see the default values in the help-files
of nls2).
They can be put in any order.
Comments begin with the '#' (dieze) character and end with the new
line character. They can be inserted anywhere.
Lower and upper case letters are not distinguished.
The logical values are noted by '0', 'no', 'non',
'false', 'faux', 'f',
or '1', 'y', 'oui', 'true', 'vrai', 't', 'v'.
Really, the first character is taken into account, only.
This file contains, in whatever order, information about
the following features of the study:
The only identifiers always required are:
- file, to describe the model,
- x and y, to describe the data,
- and theta.start, to describe the
statistical context.
- file
- pathname of the model-description-file.
- gamf
- values of the second level parameters of the
regression function.
- gamv
- values of the second level parameters of the variance function.
- vari.type
- a character value to describe the variance type among
CST, SW, VST, VB, VSB, VSTB, VTB, VI.
- eq.theta
- values of the numerical equality constraints
on the parameters Theta.
- inf.theta
- values of the inferiority-bound constraints
on the parameters Theta.
- sup.theta
- values of the superiority-bound constraints
on the parameters Theta.
- eqp.theta
- values that numbers the parameters Theta
when there are equality constraints between parameters Theta,
from 1 to the number of different parameters.
- eq.beta, inf.beta, sup.beta, eqp.beta
- same for the parameters Beta.
- Note:
if no constraint is set to a parameter,
it should be noted with NaN.
- x
- values of the independent variables, observation by observation,
the independent variables been in the same order as in the
varind declaration in the
model-description file.
- y
- values of the observed response.
- weights
- weighting values.
- curves
- character values of the curves identifiers.
Careful: The observations must be sorted, first on curves and second, on replications, if any.
- theta.start
- starting values of the parameters Theta.
- beta.start
- starting values of the parameters Beta.
- algorithm
- followed by GN or GM,
- sigma2.type
- followed by KNOWN, VARREP, VARRESID, IGNORED, or VARINTRA,
- sigma2
- its value when it is known,
- mu.type
- followed by KNOWN, MUGAUSS, MURES or MURESREP,
- mu3 and mu4
-
values of the moments when they are known
(as many values as number of observations without replications).
- max.err.c1,
max.err.c2,
max.iters,
omega.c1,
omega.c2,
lambda.start,
lambda.c1,
lambda.c2,
max.lambda,
max.stop.crit,
family,
nameN
-
other options:
see the help-pages of nls2.
-
method
- followed by values among
MLTB, MLSTB, QLTB, MLT, WLST, OLST, MLST, QLT,
VITWLS, OLSB, MLSB, QLB or MYOWN.
Put as many values as steps (maximum 3)
when the method is an alternate-method.
-
When one of method is MYOWN:
Each one of these identifiers should be followed
by as many values as methods set to MYOWN.
- nh
- the number
of the sufficient statistics Z
- effic
- a logical value
TRUE, if the estimator is efficient,
FALSE, otherwise.
- W.type
- followed by SYM, SYMBLOC, NONSYM
according to the shape of the matrix W.
- fit.type
- followed by
LOGV, STOPCRIT, NWSST, VWSS,
IVWSS, NWSSB, SIGMA2 or MYOWN.
Options can be set to control
printing,
warning,
optional results and
intermediate savings.
The identifiers are:
-
to control the print:
-
freq
- value to control the frequency of the intermediate
printings, and, also of the
intermediate savings.
- out
- the name of the printing file.
By default, it is the standard output unit.
Put "no" if no printing of any kind is requested.
Note: the upper-case letters in this
name will be automatically changed into lower-case letters,
and then, the name
should be not be equal to any of the identifiers.
-
fitted, est.eq, estim, deriv.fct,
iter, num.res, odes
-
logical values:
see wanted.print on the help-page 'nls2.control' of nls2.
to control the warnings:
- warn
- control the length of the warning messages.
It should be a value among no (or n),
long (or l),
short (or s):
n for no warning message at all,
l for long warning messages,
s for short warning messages.
- anal.warn, integ.warn, meth.warn, num.warn, all.warn
- values equal to the maximum number
of warnings allowed of types:
analyser, integrator, methodological, numerical, and
of any type.
Execution terminates
when any of these numbers is reached.
Default value: 30
- lg.trace
- the length of the
'trace', i.e the number of program-names printed
when a system error occurs
(for debugging purposes, only).
-
to control the optional results:
- as.var, correlation, b.varz.b, mu,
residuals, w
-
logical values:
F if the corresponding result is not required.
By default, all are calculated and stored.
to control the intermediate savings:
Intermediate results can be
saved along the iterating
process with a frequency controlled by
freq.
-
- it.steps
-
indexes of the steps for which
results have to be saved.
By default, it.steps is set to the number
of the last step.
it.iter, it.fitted, it.est.eq, it.estim, it.deriv.fct,
it.odes, it.num.res,
-
logical values:
T if the corresponding result is required;
see print.
Notes:
- at iteration number 0, i.e for the initial values
of the parameters, the direction,
direct, and the optimal step,omega,
are not calculated.
The corresponding returned values
are arbitraly set to DBL_MAX, (generally 1.79E+308)
a value
you find in the standard C file "float.h" on your site.
- when it.odes is set,
the fitted values of the system,
FOdes, are saved only.
Its derivatives,
d.FOdes, are not saved.
-
-
pr.odes,
itol,
atol,
rtol,
iopt,
ropt,
start,
integ.values,
cond.start,
n.theta.odes,
jt
- see the help-page
nls2.integ.ctx of nls2.
In addition here,
jt characterizes the jacobian method:
explanations can be found in
the comments included in the source-files
of LSODA.
# ****** Nasturium example ******
x # one independant variable
0.000 0.000 0.000 0.000 0.000 0.000
0.025 0.025 0.025 0.025 0.025 0.025
0.075 0.075 0.075 0.075 0.075 0.075
0.250 0.250 0.250 0.250 0.250 0.250
0.750 0.750 0.750 0.750 0.750 0.750
2.000 2.000 2.000 2.000 2.000 2.000
4.000 4.000 4.000 4.000 4.000 4.000
y # values of the response
920 889 866 930 992 1017 919 878 882 854 851 850 870 825 953
834 810 875 880 834 795 837 834 810 693 690 722 738 563 591
429 395 435 412 273 257 200 244 209 225 128 221
# the name of the model-description file:
file nas.mod
# the initial values of the parameters:
theta.start 900 -.5 1
# End of the input-file
©
INRA, 2001-