Rules

Anomaly Formula: Grubbs-Test

What it does:
Grubbs’ Test is a statistical test that detects whether the latest value is a statistical outlier in a normally distributed dataset. It checks how far the latest value deviates from the mean compared to the overall spread (standard deviation).

When to use it:
Use GrubbsTest when you want to flag extreme outliers in data that generally follows a normal distribution. It’s useful for spotting rare spikes or drops that are unlikely to happen randomly.

Examples

✅ Example 1: Increased Value Detected

• Anomaly Formula: GrubbsTest
• Change Type: Increased
• Threshold: 0.01
• History: 20, 21, 19, 22, 20, 21, 100
• Explanation:
  • Latest value = 100
  • Mean ≈ 32
  • Standard deviation ≈ 30
  • G ≈ 2.27
  • Critical value ≈ 1.90
  • Since G > critical_value → ✅ Alert triggered

❌ Example 2: Decrease Not Detected (Change Type Mismatch)

• Anomaly Formula: GrubbsTest
• Change Type: Increased
• Threshold: 0.05
• History: 40, 42, 41, 43, 40, 42, 5
• Explanation:
  • Latest value = 5 (drop)
  • G > critical → yes
  • But change type = “Increased” → ❌ No alert

✅ Example 3: Sharp Drop Caught with Change Type = Any

• Anomaly Formula: GrubbsTest
• Change Type: Any
• Threshold: 0.05
• History: 60, 61, 62, 59, 60, 12
• Explanation:
  • Latest = 12
  • Mean ≈ 52.3
  • Standard deviation ≈ 20.1
  • G ≈ 2.00, Critical ≈ 1.75
  • G > critical_value → ✅ Alert triggered