Behaviour of resampling methods under different weighting schemes, measures and variable resampling strengths

Area of Science:

• Phylogenetics
• Computational Biology
• Data Analysis

Background:

• Resampling methods are crucial for assessing the reliability of phylogenetic analyses.
• Different summarization techniques can influence the interpretation of resampling results.
• Understanding the behavior of these methods under varying resampling strengths is essential.

Purpose of the Study:

• To compare the performance of bootstrap and jackknife resampling methods.
• To evaluate different result summarization strategies (absolute frequency vs. frequency difference).
• To assess the impact of variable resampling strengths on group recovery and ranking stability.

Main Methods:

• Comparison of bootstrap and jackknife resampling techniques with variations (Poisson, symmetric).
• Evaluation using three weighting schemes and real datasets.
• Analysis of spurious groups, real groups recovered, and ranking inconsistencies using Kendall coefficient.

Main Results:

• Frequency difference (GC") generated more spurious and recovered more real groups than absolute frequency (F).
• Bootstrap methods produced more spurious groups than jackknife methods.
• Jackknife methods recovered more real groups and exhibited fewer ranking inconsistencies than bootstrap methods.

Conclusions:

• Jackknife resampling is generally more reliable than bootstrap for recovering real groups and maintaining ranking stability.
• The choice of summarization method (GC' vs. F) impacts the detection of spurious and real groups.
• Resampling strength significantly influences the outcomes of phylogenetic analyses, highlighting the importance of careful method selection.