search

trvrb / flux

Integrating influenza antigenic dynamics with molecular evolution
http://bedford.io/papers/bedford-flux/
12 stars 5 forks source link

Virus effects #31

Closed trvrb closed 11 years ago

trvrb commented 11 years ago
  • Table 1 should include a model with virus effects but no serum effects.
  • As a related point, I was surprised that inclusion of the virus effects led to only minor improvement, since prior work (http://www.ncbi.nlm.nih.gov/pubmed/19900932) suggests that some viruses are generally more resistant to HI than others. However, the virus effects are only included in a model (#9 in Table 1) where the serum effects are already included, and it is possible to imagine that including just one of these terms somehow helps correct for both effects. Can the authors include a tenth model in Table 1 that includes the virus effects but not the serum effects? This would clarify if serum effects are actually more important than virus effects, or simply whether inclusion of either one individually is more helpful than both together.
trvrb commented 11 years ago

The inclusion of 'virus effects' was meant to capture exactly this observation, that some viruses may be generally more resistant to HI. To further assess their importance, we followed this suggestion and included test error for a model that estimates virus effects, but leaves serum effects fixed at maximum titers. In this case, we observe similar, but not quite as substantial, improvements to test error as the model that estimates serum effects, but does not include virus effects. From this observation, we conclude that estimating either virus or serum effects both improve the model.