optim.fit.func: General purpose fitting function for the clanOptim library

In claus-e-andersen/clanOptim: R functions for non-linear fitting using the optim function from MASS

Description

General purpose fitting function. The actual function is supplied using the model argument. The prime use of the function is for nonlinear log-likelihood fitting.

This function (optim.fit.func) has been designed for use in connection with the mass-optimization procedure called optim(). It can be used both directly during the optimization process and before or after. Hence the function provides a general purpose wrappper such that we can easily extract anything from the Log-likelihood to prediction values for any parameter set p.

See body of function for further details. Also see the demo function.

Usage

fix.parameters(c(10,30,40),list(p5=50, p2=20)) # gives 10 20 30 40 50
  fix.parameters(NULL,list(p1=20))        # gives 20
  fix.parameters(c(10,20),list(p3=30))    # gives 10 20 30
  fix.parameters(c(10,20),list(p5=30))    # error - there is no 5th parameter
  fix.parameters(c(10,20),list(p3=30,x3=45))  # error - cant refer to the 3rd parameter more than once

Arguments

p	: a vector of parameters to be fitted. If three parameters are given
df	is the dataframe with the data to which the model will be fitted
model	: a list as shown below with information about the specific model, start-up parameters etc.
trace	: can be used to create output during each call to the function.
what:	determines what output should be returned. The default is the log-likelihood value. The following possibilities exist: 'txt' : a text string describing the function. 'name' : the name of the function. 'simulation' : a single random sample of y-values with added noise in accordanme with mu and sigma2. 'mu' : the expected values (i.e. = the model prediction if p are the fitted parameters). 'sigma2' : the variance. 'residuals' : the residials. 'residuals.std': the standardized residuals 'LogLik' : the log-likelihood value.
attach.df:	a boolean (TRUE or FALSE) indicationg if df should be attached or not. In case it is not attached thet all model$code etc. should refer to the variable inluding such as "df$dose" rather than just "dose".

Value

A merged parameter vector

Author(s)

Claus E. Andersen

claus-e-andersen/clanOptim documentation built on April 16, 2020, 5:21 p.m.