- 94 -
The other type of filtering (relative) used by VisualSpreadsheet is called Statistical filtering. Statistical
filtering uses a simple procedure where the user provides a representative training set of particles and
VisualSpreadsheet uses this set to generalize an
ideal target particle
. This target particle is then
compared against the entire data set and each particle is scored based on their ‘likeness’ or similarity to
the target.
Unlike the value filter which is strictly a boolean test for inclusion in a data subset, statistical filtering
can only rank the particles based on their ‘likeness’ to the target. The value for this likeness is
dynamically recorded for each particle as the Filter Score field. A particle's Filter Score is always a
transient value, recalculated whenever a statistical filtering operation is run against the particle. It has no
units and is not stored on disk. Filter scores lend themselves to sorting or ranking particles and so the
Filter Score field is available as a Sorting field. The default behavior of an interactive statistical filter
operation is to sort the particles in ascending Filter Score order.
A statistical filter by itself cannot generate a subset of the data. To generate a subset, a cutoff value is
needed to use with the Filter Score to partition the data into two subsets, particles like those in the
training set and particles not like those in the training set. A natural choice for the cutoff value is the
Filter Score of the highest scoring particle in the user-provided training set. This is the default value that
VisualSpreadsheet uses for this situation. Filter Score values that are less then or equal to the highest
scoring particle in the training set are considered ‘like’ the target, those with higher Filter Scores are
considered ‘unlike’. The cutoff value is simply a value filter applied after the statistical filter has been
run and its value can be manually adjusted in the filter dialog.
Filter Dialog…
The Filter Dialog is a tool that allows the user to build filters based on parameter measurements (value)
or derived statistics (statistical). The Filter Dialog can be utilized at acquisition or during post-run
analysis from the following locations (in addition to using the View Window > Filter > Filter Dialog…
pathway described in this section):
·
Main Window > Setup > Context > Filter (tab) > Advanced Acquisition Filter (section). Using a
single filter at this location allows the user to select and define which particles are going to be
captured and placed into the collage window during an analysis.
·
Main Window > Preferences > Main Window Settings > Filter Grid (tab) > Filters (section).
Using one or more filters at this location allows the user to automatically bin particles at the time
of acquisition. These filters can be displayed in the Main Window Filter Grid, with selected
statistics.
·
Main Window > Open Library or Open Classification Window. The use of the Filter Dialog in
these windows is described in the Library and Classification sections.
