CFAssay (version 1.6.0)

cellsurvLQdiff: Comparison of two linear-quadratic cell survival curves

Description

The function does an ANOVA test for overall comparison of the parameters $alpha$ and $beta$ of two linear-quadratic cell survival curves. The parameters are fitted simultaneously to the data with this function, i.e. no other function is necessary to derive the fits.

Usage

cellsurvLQdiff(X, curvevar, method="ml", PEmethod="fit")

Arguments

X

A data frame which contains columns Exp, dose, ncells, ncolonies and a further column containing two different values (character strings), which identify the two curves. Moreover, if there is no 0-value in the dose-column, X has to contain a column pe for plating efficiencies.

curvevar

Character string, which has to be one of the column names of the data frame X, that contains the two different values (character strings that distinguishes between the two curves).

method

Determines the method used for the fit. "ml" is for maximum-likelihood, "ls" for least-squares. "franken" performs weigthed least-squares with weights as described in Franken et al. (2006).

PEmethod

Controls the value of the plating efficiencies, i.e. the colony counts for untreated cells. "fit" calculates fitted plating efficiencies as model parameters, "fix" uses fixed ones calculated from the observed zero dose data or from a column named pe in X.

Value

The function returns an object of class cellsurvLQdiff containing three elements, fit1, fit2 and anv. fit1 and fit2 are objects of class glm when method="ml" or of class lm when method="ls". fit1 has parameters alpha and beta fitted in common for both cell survival curves. fit2 has parameters alpha and beta fitted differently for both curves. anv is of class anova and contains the F-test. Test results are printed, however, the full result inlcuding curve parameters is returned invisibly, i.e. the function has to be used with print or assigned to a variable, say for e.g. fitcomp as in the example below.

Details

In the data frame X, Exp identifies the experimental replicates and may be numeric or non-numeric. method="ml" for maximum-likelihood uses R function glm with family "quasipoisson" and link function "log". method="ls" uses R function lm.

See Also

glm and family with references for generalized linear modelling. anova, cellsurvLQfit.

Examples

datatab<- read.table(system.file("doc", "expl1_cellsurvcurves.txt", package="CFAssay"), header=TRUE, sep="\t")
names(datatab) #contains a column "cline"
table(datatab$cline)
fitcomp<- cellsurvLQdiff(datatab, curvevar="cline") #using default options
print(fitcomp)
plot(cellsurvLQfit(subset(datatab, cline=="okf6TERT1")), col=1)
plot(cellsurvLQfit(subset(datatab, cline=="cal33")), col=2, add=TRUE)
legend(0, 0.02, c("okf6TERT1", "cal33"), text.col=1:2)
#using different options:
print(cellsurvLQdiff(datatab, curvevar="cline", method="ls"))
print(cellsurvLQdiff(datatab, curvevar="cline", PEmethod="fix"))
print(cellsurvLQdiff(datatab, curvevar="cline", method="ls", PEmethod="fix"))
print(cellsurvLQdiff(datatab, curvevar="cline", method="franken"))