You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This bug occurred when building a tree of full-length HA sequences for the years 2000 through 2004 with frequencies weighted by region (e.g., Africa, South Asia, etc.). I noticed that the frequencies for the last pivot in 2004 did not sum to 100% but instead to ~60%. On inspection, there were no samples in this tree from Africa, South Asia, or South America. The proportion of total weight contributed by these regions is about 40%. The resulting frequency panel looked like the following:
This bug indicates that tip frequencies need to be renormalized to sum to 100% after their initial normalization to the proportions associated with each region to account for cases when not all regions are present in the tree.
The text was updated successfully, but these errors were encountered:
Fixes a bug in weighted KDE estimation of tree frequencies when one or more
weight attributes is not represented by any tips on the tree. For example, when
a tree is missing any samples from Africa but the frequencies are being weighted
by region, the total of the estimated frequencies will not sum to 1.
This commit adds unit tests to confirm the original issue and adds logic to KDE
frequency estimation to correct the issue.
Closes#268.
This bug occurred when building a tree of full-length HA sequences for the years 2000 through 2004 with frequencies weighted by region (e.g., Africa, South Asia, etc.). I noticed that the frequencies for the last pivot in 2004 did not sum to 100% but instead to ~60%. On inspection, there were no samples in this tree from Africa, South Asia, or South America. The proportion of total weight contributed by these regions is about 40%. The resulting frequency panel looked like the following:
This bug indicates that tip frequencies need to be renormalized to sum to 100% after their initial normalization to the proportions associated with each region to account for cases when not all regions are present in the tree.
The text was updated successfully, but these errors were encountered: