1999-08-09  Douglas Bates  <bates@stat.wisc.edu>

	* R/nls.R: Added summary.nls and print.summary.nls.  Cleaned up
	the definitions of nlsModel and nlsModel.plinear.

1999-06-03  Douglas Bates  <bates@stat.wisc.edu>

	* DESCRIPTION (Version): Changed version numbering to 0.2-1
	instead of 0.2.1

	* R/nls.R: Added clearNames calls in the assign calls within
	nlsModel and nlsModel.plinear to prevent the doubling of component
	names in the parameter vector.  Previously names like A.A were
	constructed instead of A.

1999-05-27  Douglas Bates  <bates@stat.wisc.edu>

	* DESCRIPTION (Package): Set up files for the 0.2.0 release.

1999-05-23  Douglas Bates  <bates@stat.wisc.edu>

	* src/nls.c: Changed the calculation of delta for the
	finite-difference derivatives.  The Hahn1 examples in the NISTnls
	library were failing in the last version.

	* R/nls.R: Added code in nlsModel and nlsModel.plinear to handle
	global variables, such as "pi" in formulas.  Previously they were
	not inserted into the environment.

	Changed instances of "e" in getAnyError to ".error" both for
	nlsModel and nlsModel.plinear

	* R/nls.R: Need to modify the code that catches the singular
	gradient matrices to set a flag that can be caught in the nls.c
	code and return the result.

1999-05-22  Douglas Bates  <bates@stat.wisc.edu>

	* R/nls.R: Need to change the names on the parameters so A.A,
	etc. are not created when just A is desired.

	In R we need to change the name of the resid.nls function to
	residuals.nls (is that consistent with S?).  Right now I have an
	as.vector in there to strip the gradient attribute.

	** must fix the model SSasympOff in zzModels.q.  It uses "input"
	where it should be "x".

1999-05-21  Douglas Bates  <bates@stat.wisc.edu>

	* R/nls.R: Changed the handling of the control argument in nls.  I
	think this way it should always have a complete set of values.

	Changed the default value of tol in nlsControl to 0.00001.  It
	seems that 0.001 was causing unfortunately premature termination
	in some cases.

	Added a default of getInitial(formula, data) for the start
	argument.  Changed the form of the mCall argument to the "initial"
	attribute from a call to a list.  Subsetting a call works
	differently in R than in S.

	Modified the order of the parameters being returned from the
	plinear form of nls to conform with the order in S.  In some ways
	it would make more sense to give the conditionally linear
	parameters first but the other convention is already established
	in S.

	Changed the name of nlsControl to nls.control to conform to the
	convention established in S.

	Will have to change the behaviour of nlsList in R so it does not
	use restart.

1999-05-20  Saikat DebRoy  <saikat@stat.wisc.edu>

	* R/nls.R: as.double() replaced with storage.mode() <- "double"
	* src/nls.c (numeric_deriv): fixed problem with increment (see
	below) by adding coercion to double in C code.

1999-05-20  Douglas Bates  <bates@stat.wisc.edu>

	* R/nls.R: Should modify nls so gradient terms are matched on
	names, not on position.

	Should add capability to profile nls models.

	Should add multiresponse estimation.

	Documentation for data sets and examples.

1999-05-19  Douglas Bates  <bates@stat.wisc.edu>

	* R/nls.R: Changed the formatting of the trace output.  Changed
	the trace output for the partially linear model to include the
	conditionally linear parameters.

	Should install the selfStart capabilities from the lme library.
	
	Need to fix the calculation of the increment and convergence
	criterion for the nls.plinear class.  The example
	nls(conc ~ cbind(1,exp(-exp(lrc)*time)), Chloride, c(lrc=-1), 
	    alg="plinear",trace = T)
	fails.
	
	* src/nls.c: Caused the trace = T option to print the initial
	state.

	* data/Nitrite.R: Change this data set to be in stacked rather
	than unstacked format.

	* data/Coal.R: Added the Coal data and all the other data sets
	from Bates and Watts (1988) "Nonlinear Regression Analysis and Its
	Applications".

	* R/nls.R: Should add a summary.nls function.
	
