[adegenet-forum] mvMapper

Julian Dupuis jrdupuis at ualberta.ca
Sat May 12 23:05:54 CEST 2018


Hi Amy,

The short answer is that the set number of hard-coded sizes (6, although
this could be configurable) is being applied to the data's range (which in
this case, I'm guessing, is 96-100%). In your case, sizing according to
that assignment support might not be the most informative, so it might be
better to try sizing with some other metadata.

For the record, here's a bit more nitty-gritty detail about how the sizing
works:

A column is treated as discrete if it is non-numeric (strings). If this is
the case, it can be assigned as the size column if the number of unique
values is less then the configured number of sizes. There are 6 sizes that
right now are hard-coded, but could theoretically be configurable if
desired. The sizes are: 6, 9, 12, 15, 18, 21.

If a column is numeric, it is treated as continuous. In this case, we first
try to bin into quantiles, again using configured number of available
sizes. Depending on the distribution, this can fail, and if quantile
binning fails, data is binned linearly across the data's range.

Hope that helps!

Cheers, Julian

--
Julian R. Dupuis, Ph.D.
Junior Faculty, University of Hawaii at Manoa
and USDA-ARS Daniel K. Inouye US PBARC
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/adegenet-forum/attachments/20180512/dff9b6d3/attachment.html>


More information about the adegenet-forum mailing list