weightEstimation: Predict weights of samples from ionomics data.
In gziegler/ionomicsUtils: Utility functions for ionomics data analysis

Description Usage Arguments Examples

Takes a data frame of values with weights included, and a data frame of values that haven't been weighed. Finds list of stable elements (based on the assumption most samples should have a similar elemental concentraiton, so low RSDs), and uses those elements to estimate weights.

1	weightEstimation(x, y, predictx, maxRSD = 25, maxEl = 10, df = 500/9)

`x`	Data frame of weight normalized elelmental concentrations.
`y`	Vector of weights for concentrations in x.
`predictx`	Data frame of values to predict weights for.
`maxRSD`	Maximum RSD threshold for elements to be included in estimation.
`maxEl`	Maximum number of elements to include in estimation.
`df`	Dilution factor for weight normalized concentrations.

##---- Should be DIRECTLY executable !! ----
##-- ==>  Define data, use random,
##--	or do  help(data=index)  for the standard data sets.

## The function is currently defined as
function (x, y, predictx, maxRSD = 25, maxEl = 10, df = 500/9) 
{
    normX <- x
    for (i in colnames(x)) {
        normX[, i] <- (x[, i]/y) * (df)
    }
    elRSD <- sapply(normX, sd, na.rm = TRUE)/sapply(normX, mean, 
        na.rm = TRUE) * 100
    elMean <- sapply(normX, mean, na.rm = TRUE)
    meetCutoff <- elRSD[which(elRSD < maxRSD)]
    if (length(meetCutoff) < maxEl) {
        numEl <- length(meetCutoff)
    }
    else {
        numEl <- maxEl
    }
    bestElements <- names(sort(meetCutoff)[1:numEl])
    weightEstimates <- predictx
    for (i in colnames(predictx)) {
        weightEstimates[, i] <- (predictx[, i]/elMean[i]) * (df)
    }
    weightEstimates <- as.data.frame(t(apply(weightEstimates[bestElements], 
        1, is.outlier, mcut = 3)))
    finalPrediction <- rowMeans(weightEstimates[bestElements], 
        na.rm = TRUE)
    return(list(weights = as.vector(finalPrediction), els = bestElements))
  }