Course contentsShow
← Data Science
Lesson 1404 of 2,14535. Statistical Anomaly DetectionFree lesson

Control Charts for Individual Observations

Monitor processes with individual measurements using I-MR charts when subgrouping isn't feasible.

Control Charts for Individual Observations

You'll learn when and how to use I-MR charts to detect process anomalies when you can't collect multiple measurements at once.

Why Individual Observations?

Sometimes you can't gather subgroups of data. Maybe your manufacturing process produces one expensive item per day, or you're monitoring monthly sales figures, or testing is destructive and costly. In these cases, traditional X-bar and R charts (which need 3-5 measurements per subgroup) won't work. That's where I-MR charts (Individual and Moving Range charts) come in.

The I-MR Chart Pair

An I-MR chart actually consists of two charts working together:

I Chart (Individuals Chart): Plots each single measurement and tracks whether the process mean is stable. Control limits are calculated as:

  • Center Line = average of all individual values
  • UCL/LCL = Center ± 2.66 × MR̄ (average moving range)

MR Chart (Moving Range Chart): Tracks variation between consecutive measurements. Each point is the absolute difference between neighboring observations. This chart confirms whether process variability is stable before trusting the I chart.

Think of it like monitoring your daily weight: you plot each day's weight (I chart) and also track how much it jumps day-to-day (MR chart). Unusual spikes in day-to-day variation warn you before the overall trend becomes obvious.

When to Use I-MR Charts

Use I-MR charts when:

  • Only one measurement per time period is available
  • Subgrouping is impossible or impractical
  • Data points are independent and reasonably normally distributed
  • You need simple, real-time monitoring

Key Takeaway: I-MR charts extend control chart methodology to scenarios with individual measurements by pairing a chart for the data itself with a chart for consecutive differences, enabling anomaly detection even without subgroups.