Analyze
The Analyze domain tests explanations for the gap identified in Define and measured in Measure. ICGB questions ask you to separate real causes from plausible stories and to interpret statistical evidence correctly.
Root-cause methods
Use fishbone diagrams, 5 Whys, process maps, and FMEA to generate candidate causes. Then verify each with data. A cause that cannot be turned into a testable statement is still a hypothesis.
Statistical reasoning
- Null hypothesis (H0): no difference or no relationship.
- p-value: probability of data this extreme if H0 is true; it is not the probability that H0 is true.
- Alpha: risk of a false alarm; beta: risk of missing a real effect; power = 1 − beta.
- Choose the test by data type: t-test for one or two means, ANOVA for three or more means, chi-square for counts, regression for relationships.
- Confidence intervals show the size of an effect, which matters more than significance alone.
Multi-vari studies
Stratify variation by position, cycle, and time to find where it enters the process.
Worked example
Two shifts produce 4.1% defects (n = 1,000) and 2.9% defects (n = 1,000). A chi-square test gives p = 0.02. At alpha = 0.05, reject H0: the rates differ. The gap is 1.2 percentage points. Reasoning: statistical significance is established, but the team must still ask whether 1.2 points is practically important and whether shift is confounded with machine or material lot. Stratify before acting.
Traps
- Treating correlation as causation.
- Testing many subgroups and reporting only the significant result.
- Ignoring normality, independence, or equal-variance assumptions.
- Calling p = 0.06 proof that no effect exists.
- Acting on practical significance without statistical support, or the reverse.