search

evogytis / fluB

Investigating the (co)evolution of reassorting influenza B lineages.
4 stars 0 forks source link

DeltaTMRCA #40

Closed trvrb closed 10 years ago

trvrb commented 10 years ago

I may be missing something obvious, but is ΔTMRCA representing "difference" or delta appropriately? If segments A and B are have differences in TMRCA that are similar to differences in TMRCA within A and within B, then ΔTMRCA is near 1. If A and B differences are bigger than ΔTMRCA approaches 0. This seems completely backwards, doesn't it? Can we flip the sign, so that completely similar is 0 and completely different is 1.

This looks like what Figure S14 (deltaTMRCAtrees) is doing? Here PB1-PB2 is low rather than high like in Figure 8 (deltaTMRCA).

evogytis commented 10 years ago

I can totally see how this might be confusing. I intended ΔTMRCA to be the actual difference in TMRCAs between two trees, which we then normalize by ΔTMRCA within each tree. I'll change the equations in Methods to reflect this, so that instead of f (A_i, B_i) in the equation it will say Δ(bar)TMRCA (A_i, B_i). As for the scale I was considering flipping the normalization equation, so that you can interpret the normalized measure as the factor by which between tree measure is bigger than the within tree measure. This would turn most comparisons to 10 (2 * between comparisons is 10 times bigger than the sum of within comparisons) and PB1-PB2-HA to something quite close to 1. Figure S14 has low values because differences in TMRCAs between pairs of tips in PB1 and PB2 trees on average do not exceed 2 years. I'll try making the nomenclature more consistent.

trvrb commented 10 years ago

I like it, but what about referring to normalized version as ΔTMRCA and the unnormalized version (used in the right side of eq 1) as δTMRCA? You can still label your plots as you have "normalized ΔTMRCA".

trvrb commented 10 years ago

Awesome!