findIPs

findIPs detects influential points (IPs) in high-dimensional data feature rankings by case deletion and quantifies their effects through rank change measurements using an adaptive weighted rank comparison approach.

Key Features:

• Influential Point Detection: Identifies cases that exert abnormal influence on feature rankings through case deletion analysis.
• Rank Change Quantification: Measures the effects of influential points by quantifying changes in feature rankings upon case removal.
• Adaptive Weighted Rank Comparison: Applies a novel rank comparing measure using adaptive weights that stress top-ranked important features and adjust to ranking properties.

Scientific Applications:

• High-Dimensional Data Analysis: Identifies individual cases that distort feature rankings in high-dimensional datasets.
• Feature Ranking Validation: Assesses the robustness and reliability of feature rankings by detecting sensitivity to individual observations.

Methodology:

The method is based on case deletion, where influential points are detected by measuring rank changes after removing individual cases, using an adaptive weighted rank comparison measure that emphasizes top-ranked features.