Z-Score
Measures how many standard deviations a value is from the mean. Fast and widely used.
✓ Use when: data is roughly bell-shaped (normal distribution) with no extreme skew.
IQR — Tukey Fences
Flags values beyond Q1 − k·IQR or Q3 + k·IQR. Doesn't depend on the mean at all.
✓ Use when: data is skewed, or you want a robust general-purpose method.
MAD — Median Absolute Deviation
Uses the median instead of the mean, making it resistant to being distorted by the very outliers you're trying to find.
✓ Use when: you suspect many outliers already exist and don't want them to inflate the threshold.
Grubbs Test
Designed to detect a single most extreme outlier at a time, iteratively. Best for small, clean datasets.
✓ Use when: you expect only 1–2 outliers and your data is otherwise well-behaved.
General Tips
→Threshold slider: Lower = stricter (flags more points). Higher = more lenient (only flags extreme values). Default 2.5 is a good starting point.
→Multivariate detection: Use the Isolation Forest section below to find outliers across multiple columns simultaneously — great for catching rows that are unusual in combination even if no single value looks extreme.
→File size limit: Best performance under 10,000 rows. Caution shown at 10k–50k rows. Blocked above 50,000 rows.
