Scatter diagrams

Key Points

• Displays the relationship between two variables using plotted data pairs.
• Helps assess correlation (positive, negative, or none) and identify outliers.
• Useful for exploring potential cause-and-effect and validating hypotheses.
• Often used in quality management, problem solving, and process improvement.
• May include a fitted trendline or correlation coefficient for clarity.
• Correlation does not prove causation; consider confounding factors.

What the Diagram Shows

• The general direction of the relationship: upward (positive), downward (negative), or no clear trend.
• The strength of association: tight cluster (strong) versus widely scattered points (weak).
• Nonlinear patterns such as curves, thresholds, or plateaus.
• Distinct clusters that may indicate subgroups or different conditions.
• Outliers that may signal data errors, special causes, or new insights.

How to Construct

• Define the purpose and choose two variables to compare.
• Collect paired data from the same observations and time frame.
• Set consistent scales and units for the x-axis (independent) and y-axis (dependent).
• Plot each data pair as a point; avoid connecting lines between points.
• Optionally add a trendline (linear or appropriate curve) and show R or R².
• Label axes, units, time period, and note any filters or subgrouping.
• Review for outliers and data quality issues; refine if needed.

Inputs Needed

• Operational definitions for both variables and expected direction of influence.
• Paired measurements from a reliable data source over a defined period.
• Units, scaling choices, and any subgroup identifiers.
• Data quality checks and a sampling plan or collection method.
• Tooling to plot points and, if needed, compute trendlines and correlation.

Outputs Produced

• A plotted chart that visualizes the relationship between two variables.
• Observed pattern: positive, negative, nonlinear, or no apparent relationship.
• Identified outliers or clusters for follow-up analysis.
• Optional trendline, regression equation, and correlation metric (e.g., R or R²).
• Insights and hypotheses to guide corrective actions or further testing.

Interpretation Tips

• Look for overall shape first, then assess strength and direction of correlation.
• Test for nonlinearity; a straight line may not fit curved patterns.
• Beware of confounders and overlapping subgroups that can mask the true relationship.
• Do not infer causation solely from correlation; validate with experiments or additional evidence.
• Check sample size and data range; restricted ranges can hide correlations.
• Investigate outliers for special causes or data errors before drawing conclusions.

Example

• A team suspects that more peer review time reduces defects. They plot review hours (x-axis) against defects found in testing (y-axis) for 25 work items. The points trend downward with a moderate fit, suggesting that increased review time is associated with fewer defects.

Pitfalls

• Using cumulative data, which can create artificial trends.
• Mixing unmatched pairs or inconsistent time frames.
• Ignoring subgroup effects that require separate plots or color-coding.
• Overreliance on R or R² without visually checking the plot for nonlinearity or outliers.
• Poor axis scaling that exaggerates or hides the pattern.
• Assuming causation and implementing changes without controlled testing.

PMP Example Question

A project team wants to verify whether an increase in training hours is related to improved first-pass quality. Which tool should they use?

1. Histogram.
2. Control chart.
3. Scatter diagram.
4. Checklist.

Correct Answer: C — Scatter diagram.

Explanation: A scatter diagram plots paired values for two variables to assess their relationship. It is the appropriate tool to explore correlation between training hours and quality outcomes.