mixtox (version 1.3)

tuneFit: Find Optimal Starting values for Curve Fitting

Description

This generic function first seaches optimal starting values based on trial and error. The the concentration-responses will be fitted using the optimal starting values. The statistics for goodness of fit is evaluated by the following statistics: coefficient of determination ($R^2$), adjusted coefficient of determination ($R_{adj}^2$), root mean squared error (RMSE), mean absolute error (MAE), Akaike information criterion (AIC), bias-corrected Akaike information criterion(AICc), and Bayesian information criterion (BIC). Thirteen sigmoidal models ("Hill", "Hill_two", "Hill_three", "Hill_four", "Weibull", "Weibull_three", "Weibull_four", "Logit", "Logit_three", "Logit_four", "BCW(Box-Cox-Weibull)", "BCL(Box-Cox-Logit)", "GL(Generalized Logit)") and four J-shaped models ("Brain_Consens", "BCV", "Biphasic", "Hill_six") are provided to fit concentration-response data.

Usage

tuneFit(conc, rspn, eq = 'Weibull', effv, highBar = 5000, bar = 1000, sav = FALSE)

Arguments

conc

a numeric vector (matrix) of experimental concentrations

rspn

a numeric vector (matrix) of responses corresponding to conc, it should have the same length (or rows and columns for matrix) as conc.

eq

models for curve fitting: "Hill", "Hill_two", "Hill_three", "Hill_four", "Weibull", "Weibull_three", "Weibull_four", "Logit", "Logit_three", "Logit_four", "BCW", "BCL", "GL", "Brain_Consens", "BCV", "Biphasic", "Hill_six".

effv

one numeric value [0, 1].

highBar

if the number of starting values exceeds highBar, A random sample of starting values will be taken.

bar

the number of random samples

sav

save fitting results in a text file (TRUE/FALSE).

Value

sta

Statistics about the goodness of fit ($R^2$, $R_{adj}^2$, MAE, RMSE, AIC, AICc, and BIC)

Details

The starting values are indispensable for nonlinear optimization algorithms. Guessing appropriate starting values requires significant experience and statistical insight. However, users in the area of toxicology always get confused and frustrated in trying so many times to fit a concentration-response curve successfully. tuneFit provides high frequency trial and error approach to find appropriate starting values for users. It will deploy those starting values one by one until finding the right one.

Examples

## example 1
# Fit the non-monotonic concentration-response data

x <- hormesis$OmimCl$x
expr <- hormesis$OmimCl$y
y <- rowMeans(expr)
tuneFit(x, y, eq = 'Biphasic')

x <- hormesis$HmimCl$x
expr <- hormesis$HmimCl$y
y <- rowMeans(expr)
tuneFit(x, y, eq = 'Biphasic', effv = 0.05)


## example 2
# Fit the concentration-response data of heavy metal Ni(2+) on MCF-7 cells.
# Calculate the concentrations that cause 5% inhibition of the growth of MCF-7

x <- cytotox$Ni$x
expr <- cytotox$Ni$y
y <- rowMeans(expr)
tuneFit(x, y, eq = 'Logit', effv = 0.05)

## example 3
# Fit the concentration-response data of Paromomycin Sulfate (PAR) on photobacteria.
# Calculate the concentrations that cause 50% inhibition of the growth of photobacteria 

x <- antibiotox$PAR$x
expr <- antibiotox$PAR$y
y <- rowMeans(expr)
tuneFit(x, y, eq = 'Logit', effv = 0.5)