FlexParamCurve (version 1.5-6)

modpar: Estimate Values to be Used for Fixed FlexParamCurve Parameters

Description

This function creates the object pnmodelparams

which holds estimates of values for all 8 FlexParamCurve

parameters used for fitting and solving positive-negative Richards curves with

SSposnegRichards and posnegRichards.eqn,

respectively.

Usage

modpar(x,
y,
pn.options = NA,
first.y = NA,
x.at.first.y = NA,
last.y = NA,
x.at.last.y = NA,
twocomponent.x = NA,
verbose = FALSE,
force8par = FALSE,
force4par = FALSE,
suppress.text = FALSE,
taper.ends = 0.45,
width.bounds = 1,
bounds.error = FALSE,
Envir = .GlobalEnv,
force.nonmonotonic = FALSE,
...)

Value

a list of estimated fixed values for all

above arguments

Arguments

x

a numeric vector of primary predictor variable

y

a numeric vector of response variable

first.y

the value of y at minimum x when it is required to be constrained

x.at.first.y

the final value of x - 0 value is used if not specified when last.y is not NA

last.y

the value of y at maximum x when it is required to be constrained

x.at.last.y

the final value of x - must be specified if last.y is not NA

twocomponent.x

a numerical specifying the x-value (e.g. age) of intersection if a double model of

two separate components is to be fitted. Alternatively a logical of value

= TRUE if the same type of model is to be fitted but the x of

intersection is unknown

verbose

logical indicating whether information on successful optimization and

parameters should be returned during when using SSposnegRichards

force8par

logical specifying whether parameters of the negative Richards

curve should be set to defaults if they cannot be estimated

force4par

logical specifying whether parameters of the negative Richards

should be ignored - effectively using simple Richards curve

pn.options

character string specifying name of list object populated with starting

parameter estimates, fitting options and bounds or the destination for modpar to write a new list

suppress.text

logical specifying whether modpar should return descriptive text to the screen during execution

taper.ends

numeric representing the proportion of the range of the x variable for which data are extended at

the two ends of the data set. This is used in initial estimation (prior to optim and nls optimizations) and can

speed up subsequent optimizations by imposing a more pronounced S-shape to both first and second curves. Defaults to 0.45.

width.bounds

a numeric indicating the proportion of the usual width of parameter bounds to be imposed during optimizations.

Large values may slow or terminate computations, however they could better accomodate data in which different levels exhibit very different

parameter values.

bounds.error

a logical. If true parameter estimation will terminate if initial estimation of parameters leads to

values outside specified bounds in pn.options. If false, more appropriate bounds will be determined automatically.

Envir

a valid R environment to find pn.options in and export any output to, by default this is the global environment

force.nonmonotonic

if set to TRUE fixed recessional parameter estimates will be used for the two monotonic equations

(modno #12 or #32), otherwise these two models will use RAsym = 0, Ri = 0, Rk = 1, RM = 1 to prevent non-monotonic relationships

in these cases.

...

additional optional arguments to be passed to nlsList

Author

Stephen Oswald <[e-mail]>

Details

This function creates a formatted list object as named by the argument pn.options. This list

holds estimates of values for all 8 FlexParamCurve parameters, fitting options and parameter bounds used for

fitting and solving double-Richards curves with SSposnegRichards and posnegRichards.eqn,

respectively. Parameter bounds are the maximum and minimum parameters values that can be used by optim

and nls during parameter estimation. For definitions of parameters see either SSposnegRichards

or posnegRichards.eqn. The list is written automatically by the function (to ".pntemplist") but it is

also output as a return value for assignation and subsequent use in the usual way [myoptions<- change.pnparameters(...)].

Estimates are produced by fitting positive-negative or double Richards curves in

nls using

SSposnegRichards for the full 8 parameter model (R1).

If this fails, the function getInitial is called to

attempt to produce initial estimates using the same 8 parameter model.

If this also fails, estimates are attempted in the same way using the

4 parameter (positive only) model (R12). In this case, only the positive

parameters are returned (NAs are substituted for negative parameters)

unless argument force8par=TRUE, in which case negative parameters are

defaulted to: RAsym = 0.05*Asym, Rk = K, Ri = Infl, RM = M.

This function can now fit biphasic (and more generally

double-Richards) curves, where the final curve is effectively either two positive curves

or two negative curves, as well as negative-positive curves. This functionality is default

and does not need to be specified.

Parameter bounds estimated here for use in optim and nls

fits within SSposnegRichards are

applicable to a wide range of curves, although user may

change these manually in list object specified by pn.options.

Bounds are estimated by modpar by adding or subtracting multiples

of fixed parameter values to estimated mean parameter values:

-Asym*0.5 and +Asym*2.5,

-K*0.5 and +K*0.5,

-Infl*2.5 and +Infl*10

-M*2 and +M*2

-RAsym*0.5 and +RAsym*2.5,

-Rk*0.5 and +Rk*0.5,

-Ri*2.5 and +Ri*5

-RM*2 and +RM*2.

Use force8par = TRUE if initial call to modpar produces estimates for

only 4 parameters and yet an 8 parameter model is desired for SSposnegRichards

or posnegRichards.eqn.

Use force4par = TRUE if you desire to produces estimates only for the four parameters of

a single Richards curves. This should also be used if you wish to fit simple logistic

Gompertz or von Bertalanffy curves: see SSposnegRichards for more details. If

the specified model in subsequent \(SSposnegRichards\), model selection or ploting calls

is monotonic (i.e. contains no recession parameters: modno= 12 or 32) recessional parameters

will not be included for these two models unless "force.nonmonotonic" option is TRUE,

in the specified pn.options list object, in which case parameters will be drawn from the specified

pn.options list object.

When specified, first.y and last.y are saved in list object specified by pn.options to instruct

SSposnegRichards to add this as the first or last value of the response, respectively,

during estimation.

To fit two-component double-curves, in which one curve equation is used up to (and including)

the x of intersection and a separate equation is used for x-values greater than the x of intersection

the argument twocomponent.x should be set to the value for the x of intersection. If this argument

is anything other than NA then a two-component model will be fitted when SSposnegRichards

is called. This option will be saved in list object specified by pn.options and can be changed at will.

taper.ends can be used to speed up optimization as it extends the dataset at maximum and minimum extremes

of x by repeatedly pasting the y values at these extremes for a specified proportion of the range of x.

taper.ends is a numeric value representing the proportion of the range of x values are extended for and

defaults to 0.45 (45

tend towards a zero slope this is a suitable values. If tapered ends are not desirable then choose taper.ends = 0.

If the argument verbose = TRUE then details concerning the optimization processes within

SSposnegRichards are printed on screen whenever SSposnegRichards is called.

These include whether optimization of the first or second parts of the curve or simultaneous optimizations

are successful, if these have been further refined by nls, whether default parameters were used or the

parameterization was aborted and what parameter values were finally exported by SSposnegRichards.

This option will be saved in the list object specified by pn.options and can be changed at will.

Examples

# estimate fixed parameters use data object posneg.data

   	 modpar(posneg.data$age, posneg.data$mass, pn.options = "myoptions")



# estimate fixed parameters use data object posneg.data (only first 

# 4 group levels for example's sake) and specify a fixed hatching 

# mass for curve optimization using \code{\link{SSposnegRichards}}

	 modpar(posneg.data$age, posneg.data$mass, pn.options = "myoptions")

   	 subdata <- subset(posneg.data,posneg.data$id == as.character(36)

	    		| posneg.data$id == as.character(9) 

	    		| posneg.data$id == as.character(32) 

	    		| posneg.data$id == as.character(43))

	 richardsR22.lis <- nlsList(mass ~ SSposnegRichards(age, Asym = Asym, 

	         K = K, Infl = Infl, RAsym = RAsym, Rk = Rk, Ri = Ri, 

        	 modno = 22, pn.options = "myoptions"), data = subdata)

   		 



# force an 8 parameter estimate on logistic data

modpar(logist.data$age,logist.data$mass,force8par=TRUE, pn.options = "myoptions")





# force an 4 parameter model on logistic data

modpar(logist.data$age,logist.data$mass,force4par=TRUE, pn.options = "myoptions")



# troubleshoot the fit of a model 

modpar(posneg.data$age,posneg.data$mass,verbose=TRUE, pn.options = "myoptions")



# fit a two component model - enter your own data in place of "mydata"

        # this details an approach but is not run for want of appropriate data

        # if x of intersection unknown

        if (FALSE) {

 	modpar(mydata$x,mydata$y,twocomponent.x=TRUE, pn.options = "myoptions")

        # if x of intersection = 75

 	modpar(mydata$x,mydata$y,twocomponent.x=75, pn.options = "myoptions")

 	richardsR1.nls <- nls(y~ SSposnegRichards(x, Asym = Asym, K = K,

                 Infl = Infl, M = M, RAsym = RAsym, Rk = Rk, Ri = Ri, RM = RM,

                 modno = 1, pn.options = "myoptions")

                 , data = mydata)}