Certified Six Sigma Black Belt Advanced Hypothesis Testing Questions and Answers — Questions and Answers
Question 1: A Six Sigma Black Belt is analyzing customer satisfaction scores (on a 1-10 scale) for three different service centers (A, B, C). The data is not normally distributed, and a test for equal variances has failed. Which non-parametric test is most appropriate to determine if there is a statistically significant difference in the median satisfaction scores among the three centers?
- ANOVA
- Mann-Whitney U Test
- Kruskal-Wallis Test (Correct answer)
- 1-sample Sign Test
Correct answer: Kruskal-Wallis Test
The Kruskal-Wallis Test is the correct choice because it is a non-parametric alternative to a one-way ANOVA. It is used to compare the medians of two or more independent groups when the data is not normally distributed. ANOVA is a parametric test and requires assumptions of normality and equal variances, which are violated here. The Mann-Whitney U Test is used for only two groups, and the 1-sample Sign Test is for comparing the median of a single group to a target value.
Question 2: A project team is investigating the relationship between the type of raw material supplier (four suppliers, a categorical variable) and the occurrence of a specific defect (present/absent, a categorical variable). Which statistical test should be used to determine if there is a significant association between the supplier and the defect rate?
- 2-sample t-test
- Chi-Square Test of Independence (Correct answer)
- Mood's Median Test
- Simple Linear Regression
Correct answer: Chi-Square Test of Independence
The Chi-Square Test of Independence is used to determine if there is a significant association between two categorical variables. In this scenario, 'supplier' and 'defect occurrence' are both categorical. A t-test compares the means of continuous data, Mood's Median test compares medians, and regression analyzes the relationship between continuous variables.
Question 3: When performing an Analysis of Variance (ANOVA), a key assumption is the homogeneity of variances. If a Black Belt suspects the data may not be normally distributed, which test is more robust for verifying the equality of variances among the groups?
- Bartlett's Test
- Anderson-Darling Test
- Levene's Test (Correct answer)
- Shapiro-Wilk Test
Correct answer: Levene's Test
Levene's Test is more robust than Bartlett's Test when the data deviates from a normal distribution. Bartlett's Test is highly sensitive to non-normality and can produce misleading results in such cases. The Anderson-Darling and Shapiro-Wilk tests are used to check for normality, not for equality of variances.
Question 4: A Black Belt wants to analyze the effect of a new training program (a categorical factor with two levels: 'New' and 'Old') on employee productivity (a continuous variable). However, they know that an employee's prior experience (a continuous variable, measured in months) also significantly impacts productivity. To isolate the effect of the training program while accounting for the influence of prior experience, which statistical technique is most appropriate?
- Multiple Regression
- Two-Way ANOVA
- Analysis of Covariance (ANCOVA) (Correct answer)
- Mann-Whitney U Test
Correct answer: Analysis of Covariance (ANCOVA)
Analysis of Covariance (ANCOVA) is the ideal technique. It combines elements of ANOVA and regression to compare the means of groups (the training programs) while statistically controlling for the effect of a continuous variable (the covariate, which is prior experience). This allows for a more accurate assessment of the training program's true effect. Multiple regression could be used, but ANCOVA is specifically designed for this type of experimental question with a mix of categorical factors and continuous covariates.
Question 5: In a multiple regression analysis, a Six Sigma team is concerned about multicollinearity. Which of the following is the BEST indicator that multicollinearity is a problem in their model?
- A low R-squared value.
- A high p-value for the overall F-test.
- Residuals that are not normally distributed.
- High Variance Inflation Factor (VIF) values for predictor variables. (Correct answer)
Correct answer: High Variance Inflation Factor (VIF) values for predictor variables.
A high Variance Inflation Factor (VIF), typically greater than 5 or 10, is the primary diagnostic for multicollinearity. Multicollinearity occurs when predictor variables in a regression model are highly correlated with each other, which can inflate the variance of the coefficient estimates and make the model unstable. A low R-squared suggests the model doesn't explain much variance, a high p-value for the F-test suggests the model is not significant, and non-normal residuals violate a key assumption of regression but do not indicate multicollinearity.
Question 6: Which of the following scenarios is the most appropriate application for a non-parametric hypothesis test?
- Comparing the means of two large, normally distributed samples with equal variances.
- Analyzing the relationship between two continuous variables that show a strong linear pattern.
- Comparing the median cycle times of three production lines where the data for each line is highly skewed. (Correct answer)
- Testing if the variance of a normally distributed process output is equal to a target value.
Correct answer: Comparing the median cycle times of three production lines where the data for each line is highly skewed.
Non-parametric tests are used when the assumptions of parametric tests, such as normality, are not met. Skewed data is a classic example of non-normal data. Therefore, comparing the medians (a measure of central tendency used in non-parametric tests) of three groups with skewed data is the perfect application. The other options describe situations where parametric tests like the t-test, linear regression, or chi-square test for variance would be appropriate.
Certified Six Sigma Black Belt Advanced Hypothesis Testing Questions and Answers
A Six Sigma Black Belt is analyzing customer satisfaction scores (on a 1-10 scale) for three different service centers (A, B, C).
The data is not normally distributed, and a test for equal variances has failed.
Which non-parametric test is most appropriate to determine if there is a statistically significant difference in the median satisfaction scores among the three centers?