An outlier calling method that shrinks the observed variance for a bead-type towards the predicted variance based on all bead-types on the array-section.

1 | ```
squeezedVarOutlierMethod(inten, probeList, wts=1, n = 3, predictNlim=14)
``` |

`inten` |
a list of intensities |

`probeList` |
the IDs corresponding to each intensity value |

`wts` |
Weights associated with beads, indicating those recommended for removal by, for example, |

`n` |
number of SDs cutoff used |

`predictNlim` |
how many beads of a bead-type must be present for that bead-type to contribute to prediction of variances? |

This function is called within the `summarize`

routine of beadarray to exclude outlying beads from an array-section prior to summary. The intensities are not assumed to be on any particular scale and can result from any user-defined transformation function, however a log-transformation is recommended.

Bead-types that have `predictNlim`

numbers are used to locally regress bead-type precision against bead-type mean, as well as the squared residual error of bead-type precision against bead-type mean. These are then used as prior values for the distribution of precision to feed into a standard Bayesian calcuation to obtain an estimate of the posterior variance.

Beads with weight zero do not contribute to the outlier calling.

the positions in the original vector that were determined to be outliers

Andy Lynch

`illuminaOutlierMethod`

1 2 3 4 5 6 7 | ```
if(require(beadarrayExampleData)){
data(exampleBLData)
oList = squeezedVarOutlierMethod(logGreenChannelTransform(exampleBLData, 1), getBeadData(exampleBLData, array=1, what="ProbeID"))
}
``` |

Questions? Problems? Suggestions? Tweet to @rdrrHQ or email at ian@mutexlabs.com.

All documentation is copyright its authors; we didn't write any of that.