BoxWhisker Filter

This filter provides a method of identifying outlier probes based on their position in the whole distribution of probe values. It is based on a BoxWhisker plot and is therefore applicable to any kind of distribution of probe values.

Options

• You need to select one or more DataStores from the list on the left. Data Sets are shown in red and Data Stores are shown in Blue.
• You need to select a stringency. The formula for picking a cutoff for outliers is the median +/- the interquartile range x the stringency. A stringency of 1.5 is usual for defining outliers. A stringency of 3 is often used for distant outliers, but you can pick any value you feel to be appropriate
• If you have selected more than one DataStore for the filter you need to define in how many stores a probe needs to be an outlier before it goes into the filtered set. You can express this using an Exact, At least or No More Than cutoff.

For a visual representation of the outliers in your dataset you can use the BoxWhisker Plot. This shows a BoxWhisker plot with a stringency of 2.