search

hyanwong / treeseq-node-identifiability

Examples and issues to assess identifiability of nodes in an arg / tree sequence
MIT License
0 stars 0 forks source link

defining identifiable nodes #3

Open GertjanBisschop opened 1 year ago

GertjanBisschop commented 1 year ago

This will be mostly useful for myself. Definition: A binary node is identifiable when there is sufficient information to define its two children (when we say such a node is 2-identifiable, can this be extended/relaxed to n-identifiable?).

The nodes we want to identify are those nodes that have multiple parents, meaning, the first coalescent nodes on the path descending from a recombination event. Here we should have more information at our disposal then for all other nodes. For all other nodes: the criterium is clear: we require at least one mutation on the edge in the ARG that has this node as its child.

GertjanBisschop commented 1 year ago

We defined a rough outline for a potential way to find identifiable nodes for the case when there are no recombination events for the paths from our focal node to the leaves. However, we need a good, yet simple example for the case when there is such a recombination event 'underneath' our focal node.