Plot slope ratios on a log scale

Plot the target-absent to target-present slope ratio with a log transform instead of raw ratio values. For example, use a histogram of log(slope ratio) or summarize ratio bins with mean log(ratio) or median ratio when raw ratios show a long right tail.

Raw slope ratios expand sharply when the target-present slope gets close to zero, so a few efficient searches can dominate the display. A log transform compresses those extremes and makes the ratio distribution much closer to normal.

Mechanism: Log encoding reduces the visual and statistical pull of extreme small-denominator ratios, so the plotted distribution better reflects the typical relationship between target-present and target-absent slopes.

Evidence: The paper reports that untransformed slope ratios were strongly positively skewed, that log-transforming them made the distribution roughly normal, and that ratios became especially large as target-present slopes approached zero (Wolfe, 1998).

Notes: The paper also plots median ratio and mean log(ratio) across target-present slope bins to reduce the effect of small denominators.

• User Goal: Compare slope-ratio distributions or judge whether ratios differ from a 2:1 benchmark.
• Task: Summarize or visualize target-absent to target-present slope ratios across many observations or slope bins.
• Data: Quantitative slope pairs with some very small target-present slopes and a positively skewed raw ratio distribution.
• Chart Setting: A ratio histogram or a binned summary plot of slope ratio.
• Audience: Domain experts reading visual-search results.
• Success Criterion: The ratio view is not dominated by a few extreme values from very small denominators.

Break it when: Nonpositive ratios must be shown unchanged, or target-present slopes near zero cannot be filtered or handled separately. Why: Log values are undefined for nonpositive ratios, and near-zero denominators make raw ratios unstable.

Sacrifice: Raw ratio magnitudes become less direct to read off the axis. Risk: Readers can compare logged values to a raw 2.0 benchmark if the transform is not stated clearly. Mitigation: State that the axis or summary uses log(slope ratio) and note any removed nonpositive cases.

Mistake: Plot raw mean slope ratios across all observations without handling small target-present slopes. Why it fails: A small number of near-zero denominators can stretch the display and distort the apparent center of the distribution.

Failure Sign: The ratio plot has a long right tail or a few extremely large values. Quick Check: Count nonpositive ratios and inspect whether many target-present slopes are close to zero before plotting raw ratios. Stronger Test: Compare the raw-ratio distribution with a log-ratio version and confirm that the log-transformed view is substantially less skewed.

• Remove nonpositive ratio cases before plotting log(slope ratio).
• Replace a raw-ratio axis with a log-ratio axis for the main distribution view.
• Report median ratio or mean log(ratio) in slope bins where small target-present slopes inflate raw means.