getForegroundSD.SpotData: Gets (an approximation of) the standard deviation of the...
In HenrikBengtsson/aroma: An R Object-oriented Microarray Analysis package

Description Usage Details Value Author(s) See Also Examples

Gets (an approximation of) the standard deviation of the foreground pixels. Since the Spot software does not return estimates of the standard deviation of the foreground pixels, but the interquartile range (IQR) of their logarithm instead, the value returned by this method is only an approximation of the standard deviation. It is based on the assumption that the noise is symmetric on the non-logaritmic scale.

1 2	## S3 method for class 'SpotData' getForegroundSD(this, ...)

By assuming that the noise is symmetric on the non-logaritmic scale we can first derive the IQR on the non-logaritmic scale since we know the median pixel value too. Let the non-logaritmic pixel value be X and the let x=log[2](X). More over, let X[0.50] be the median value of all X's, X[0.25] and X[0.75] be the 0.25 and 0.75 quantile, respectively. x[0.25], x[0.50], x[0.75] are defined analogously. Assuming symmetric noise on the non-log scale means that X[0.50]-X[0.25] == X[0.75]-X[0.50] is true. Let dX = X[0.75]-X[0.50]. Given the IQR or the log pixel values, i.e. xIQR=x[0.75]-x[0.25], one can derive

dX = (2^xIQR-1/(2^xIQR+1)*X[0.50]

Hence, the IQR of the non-logarithmic pixel values can then be approximated by 2*dX. If the noise is not only symmetric, but also Gaussian, the standard deviation is 1.35*2*dX, which is the value returned.

Returns a list of matrices that contain approximations of the estimated standard deviation of the pixels in the foreground region of the spots.

Henrik Bengtsson (http://www.braju.com/R/)

For more information see SpotData.

  spot <- SpotData$read("spot123.spot", path=system.file("data-ex", package="aroma"))
  raw <- getRawData(spot)
  sd <- getForegroundSD(spot)
  raw$RSD <- sd$RSD; raw$GSD <- sd$GSD;
  subplots(4)
  plot(raw, "RSDvsR", col="red")
  plot(raw, "GSDvsG", col="green")