averageData: Average biological replicates over different slides.

Description

The function assumes that each signal originates from an underlying true value which is scaled by a scaling factor depending on the slide and replicate. The method optimizes the scaling and truth parameters such that the distance between predicted and actual signals is minimized. There are aguments to specify what factors the scaling factors and truth parameters depend on.

Usage

averageData(subsample, scaling = c("slide", "replicate"), 
	distinguish = c("cellline", "treatment"))

Arguments

subsample

data.frame with columns "slide" (factor, the slide names), "ab" (factor, the antibody/target names), "time" (numeric, the time points), "signal" (numeric, signal values), "sigma0" (numeric, error parameter for the constant error), "sigmaR" (numeric, error

scaling

character. One scaling parameter ist estimated for each occurring combination of the corresponding factors.

distinguish

character. One truth parameter ist estimaed for each occuring combination of the factors "time", "ab" (antibody/target) and the factors in distinguish.

Value

data.framewith columns "time", "ab", "signal" (the truth parameters returned by nls), "sigma" (the standard error of the truth parameter returned by nls), "connection" (integer, signals can only be compared on the same scale if they agree in "ab", and "connection") and one column for each entry of distinguish.

Details

Averaging is based on the assumption that for each level of scaling there is an underlying "true" antibody time-course for each level of distinguish. The signals of different scaling levels are assumed to differ by a scaling factor. Both, antibody time-course values and scaling parameters are estimated simulatenously by generalized least squares estimation: $$GRSS = \sum_{i,j}(s_i S_{ij}- y_j/s_i)^2/(\sigma_{ij,0}^2 + (y_j/s_i)^2 \sigma_{ij,R}^2)$$ where $i, j$ correspond to the levels of c("time", "ab", distinguish) and the levels of scaling.