Course contentsShow
← Data Science
Lesson 1413 of 2,14535. Statistical Anomaly DetectionFree lesson

Types of Changes: Mean, Variance, and Trend

Distinguishing between changes in level, variability, and slope, and how each manifests in time series data.

Types of Changes: Mean, Variance, and Trend

You'll learn how to distinguish the three fundamental types of changes that occur in time series data: shifts in level, variability, and slope.

Understanding the Three Types of Changes

When detecting change-points in time series, not all changes look the same. Think of monitoring a river: sometimes the water level suddenly rises (mean shift), sometimes it becomes choppier with bigger waves (variance change), or sometimes it gradually rises over hours (trend change).

Mean (Level) Changes

A mean change occurs when the average value of your data shifts up or down but maintains similar variability. Imagine factory output suddenly jumping from 100 units/day to 150 units/day after installing new equipment—the level changed, but day-to-day fluctuations remain consistent.

Variance (Spread) Changes

A variance change affects how much the data fluctuates around its mean, without necessarily changing the average itself. Picture website traffic: your daily average might stay at 10,000 visitors, but suddenly it becomes unpredictable—swinging between 5,000 and 15,000 instead of the usual 9,000 to 11,000 range.

Trend (Slope) Changes

A trend change involves a shift in the rate of increase or decrease. Think of sales that were flat for months suddenly growing 5% per week, or growth that was accelerating suddenly slowing down. The direction or speed of movement changes.

Why the Distinction Matters

Each type of change requires different detection techniques and has different business implications. A mean shift might indicate a process improvement, variance increase could signal instability, and trend changes often reflect market dynamics or seasonal patterns.

Key Takeaway: Time series changes manifest in three primary ways—shifts in average level (mean), fluctuation magnitude (variance), or directional movement (trend)—and recognizing which type occurred is essential for proper anomaly interpretation and response.