2-Sample T-Test in Minitab – Tabtrainer® Quality Benchmark

In this training session, participants will explore the application of the two-sample t-test to evaluate whether the population means of two independent groups differ significantly. Using a practical scenario from the Smartboard Company, the course focuses on quality control and supplier evaluation for high-strength screws used in skateboard assembly for the Olympic team.

Key learning content includes:

1. Understanding Mechanical Properties:

   • Explanation of tensile strength and yield strength for 10.9-grade screws.

   • The importance of yield strength in ensuring elasticity and preventing permanent deformation under high loads.

2. Supplier Performance Context:

   • Challenges faced by Smartboard Company due to inconsistent deliveries from their primary supplier (Winters).

   • Introduction of a new supplier (Ziegler) to meet increased demand.

   • Ensuring no qualitative differences between screws from the two suppliers.

3. Statistical Benchmarking Goals:

   • Defining performance benchmarks for screw properties (mean yield strength ≥ 915 MPa).

   • Determining statistical significance of differences using a threshold of 1.5 times the standard deviation.

4. Sample Size Determination:

   • Conducting a hypothesis power analysis to calculate the required sample size for achieving 90% power in the statistical test.

The session integrates practical statistical tools with real-world applications, equipping participants with skills to analyze data, compare supplier quality, and provide actionable recommendations to ensure consistent product performance.

In this section of the training, participants will learn how to determine the optimal sample size required for a two-sample t-test to achieve a specified hypothesis power. Using the Smartboard Company scenario, the focus is on ensuring robust statistical conclusions about the yield strength of screws from two suppliers. Key topics covered include:

1. Hypothesis Power and Sample Size Determination:

   • The importance of achieving a hypothesis power of 80-90%, with 90% being the standard for Smartboard Company.

   • Using statistical software to calculate the required sample size based on power, difference, and standard deviation inputs.

2. Step-by-Step Power Analysis:

   • Selecting the "Two-Sample T-Test" option in statistical tools.

   • Inputting power values (e.g., 90%, 95%, and 99%) and defining detectable differences (e.g., 1.5 times the standard deviation).

   • Averaging historical standard deviations from both suppliers to calculate a replacement standard deviation.

3. Interpreting Results and Setting Parameters:

   • Understanding why sample sizes are rounded to whole numbers to ensure practical applicability.

   • Observing that the required sample size remains consistent regardless of whether the difference is positive or negative.

4. Application to Real Data:

   • Determining a sample size of 20 for 90% power and rounding up to 25 to ensure greater reliability.

   • Importing and inspecting sample data from suppliers into statistical software to confirm data completeness.

5. Practical Insights:

   • Recognizing that proper sample size calculations improve test reliability and minimize errors.

   • Preparing the dataset for further analysis, such as graphical and statistical comparisons.

This section equips participants with the tools and knowledge to calculate accurate sample sizes for hypothesis testing, ensuring reliable and actionable insights in quality control and supplier evaluations.

In this section, participants will learn how to visualize and interpret data using dot plots, focusing on comparing two data sets for yield strength values. The content includes practical guidance on creating and refining visualizations to identify trends and outliers. Key points covered are:

1. Choosing the Right Visualization:

   • Dot plots are recommended for datasets with fewer than 50 values.

   • Histograms are suitable for larger datasets with 20 or more values.

   • In the Smartboard scenario, dot plots are selected to compare yield strength distributions for screws from suppliers Winters and Ziegler.

2. Dot Plot Types and Selection Criteria:

   • Explanation of different dot plot types such as One Y, Simple, One Y, With Groups, and Multiple Y’s.

   • Criteria for selecting the appropriate format based on dataset structure (e.g., grouped or ungrouped data).

   • For this analysis, the "Multiple Y’s Simple" format is chosen to compare the two groups in separate plots.

3. Handling Outliers and Data Preparation:

   • Identification of outliers (e.g., zero values caused by measurement errors).

   • Replacing missing or erroneous values with asterisks to prevent distortion in statistical analysis.

   • Updating results after data correction to ensure accurate visualizations.

4. Interpreting the Dot Plot Results:

   • Axis scaling adjustments for better data clarity and resolution.

   • Observing data trends, such as greater scatter in the yield strength values for Winters compared to Ziegler.

5. Practical Steps for Creating and Refining Dot Plots:

   • Using software tools to create dot plots, customize scaling, and update visualizations.

   • Ensuring the plots provide clear and actionable insights for identifying differences in data groups.

This section equips participants with the skills to effectively visualize and analyze data distributions, identify potential issues (e.g., outliers), and prepare accurate inputs for subsequent hypothesis testing.

This section focuses on advancing the visual and statistical analysis of data using boxplots and graphical summaries. Participants will learn to supplement dot plot insights with more robust statistical tools to identify trends and assess data distributions effectively. Key highlights include:

1. Caution with Graphical Observations:

   • Visual trends in yield strength and data scattering (e.g., greater variability in Winters' screws) must not be overinterpreted as they might arise by chance.

   • Statistical tests are necessary to validate whether observed differences are significant at the population level.

2. Introduction to Boxplots for Data Comparison:

   • Boxplots provide a clear view of data spread, median values, and potential differences in arithmetic mean.

   • Customizing boxplots to display mean values alongside medians enhances interpretability.

   • Adjusting the visual layout ensures clarity for data-driven decision-making.

3. Insights from Boxplot Analysis:

   • The Winters sample shows higher arithmetic mean yield strength but with greater data scatter.

   • Ziegler’s screws exhibit tighter clustering, indicating more consistent quality.

   • These visual trends require further statistical confirmation to rule out randomness.

4. Graphical Summary for Statistical Insights:

   • Combining dot plots, boxplots, and graphical summaries offers a comprehensive analysis.

   • Summaries include location and dispersion parameters (mean, standard deviation, variance, skewness, kurtosis) and confidence intervals.

5. Normality Testing with Anderson-Darling:

   • Both samples pass the normality test with p-values greater than 0.05, confirming adherence to the null hypothesis of normal distribution.

   • Skewness and kurtosis values provide further insights into data symmetry and spread compared to an ideal normal distribution.

6. Practical Steps for Creating Graphical Summaries:

   • Using software tools to generate comprehensive statistical displays.

   • Inputting variables appropriately to account for sample structure (e.g., separate columns for Winters and Ziegler).

   • Interpreting confidence intervals to understand population-level implications of sample data.

This segment ensures participants can effectively combine graphical and statistical tools to analyze sample data and prepare for hypothesis testing, reinforcing the reliability of their insights and recommendations.

This section focuses on the application of statistical testing, particularly the two-sample t-test, to evaluate differences in mean values and distribution characteristics. Participants will learn to interpret statistical results and derive meaningful conclusions. Key topics include:

1. Understanding Kurtosis and Skewness:

   • Negative kurtosis values for both samples indicate higher data density in the middle range compared to the ideal normal distribution, with weaker density in the tails.

   • Skewness values approaching zero signify symmetric distributions, essential for accurate hypothesis testing.

1. Quartiles and Confidence Intervals:

   • Quartiles divide datasets into meaningful sections (e.g., 25% below Q1, 50% between Q1 and Q3).

   • Confidence intervals allow indirect statistical conclusions about populations based on samples:

     • Example: 95% confidence that the mean yield strength of Winters screws lies between 914.45 MPa and 916.09 MPa.

1. Importance of Comprehensive Statistical Parameters:

   • In non-ideal normal distributions, multiple location and dispersion parameters (e.g., median, quartiles, standard deviation) provide robust insights.

   • The arithmetic mean is sensitive to outliers, while the median offers stability in asymmetric distributions.

1. Performing the Two-Sample T-Test:

   • Purpose: Test whether the mean values of two independent samples (Winters vs. Ziegler) are statistically different.

   • Null Hypothesis (H₀): No difference exists between the means.

   • Alternative Hypothesis (H₁): A statistically significant difference exists.

1. Key Steps in Executing the T-Test:

   • Importing data and assigning samples into separate columns.

   • Setting a 95% confidence level for testing.

   • Inputting the null hypothesis value (mean difference = 0).

   • Keeping the "Assume Equal Variances" option inactive to account for unknown variances.

1. Interpreting T-Test Results:

   • Sample sizes: 24 screws per supplier.

   • Mean values: Winters = 915.27 MPa, Ziegler = 913.58 MPa.

   • Standard deviations: Winters = 1.94 MPa, Ziegler = 0.985 MPa.

   • Calculated p-value: If below 0.05, reject the null hypothesis and accept the alternative hypothesis, indicating significant differences between suppliers.

1. Practical Application of Results:

   • A statistically significant difference in mean yield strength indicates potential qualitative differences between suppliers.

   • The analysis enables Smartboard Company to evaluate and benchmark suppliers' performance accurately.

This section provides a detailed exploration of statistical methods for validating differences in yield strength and variance between two suppliers, Winters and Ziegler. Participants will learn to assess statistical significance, interpret results, and formulate actionable recommendations based on data analysis.

1. Statistical Confirmation of Differences

• T-Test Results:

  • A t-value of 3.8 and a p-value of 0.001 indicate statistically significant differences between the mean yield strengths of the two suppliers.

  • Null Hypothesis (H₀): Rejected as the p-value < 0.05, confirming the observed differences are not due to chance.

  • Alternative Hypothesis (H₁): Accepted, verifying meaningful differences between the arithmetic means of the two populations.

• Degrees of Freedom (DF):

  • Initially 48 (24 samples each from Winters and Ziegler).

  • Calculations considering unequal variances use 14 DF, which reflects the data available for parameter estimation.

2. Variance Analysis and Process Stability

• Variance Test Results:

  • Variances for Winters and Ziegler are 1.936 and 0.985, respectively, with a variance ratio of 1.964.

  • P-values (Bonett: 0.003, Levene: 0.005) confirm statistically significant differences in variance.

  • Conclusion: Winters shows higher yield strength but greater process variation compared to Ziegler, indicating less process control.

3. Interpretation and Insights

• Winters:

  • Strength: Higher average yield strength.

  • Weakness: Greater variability and inconsistency in process stability.

• Ziegler:

  • Strength: Consistent and reliable quality due to lower process variance.

  • Weakness: Lower yield strength compared to Winters.

4. Recommendations for Action

• Winters:

  • Improve process stability by employing tools such as quality control charts and process capability analyses.

  • Identify and adjust variables causing variation using design of experiments or correlation and regression analysis.

• Ziegler:

  • Enhance yield strength through optimization of forming and heat treatment parameters.

5. Presentation of Results

• Layout Tool for Reporting:

  • Use graphical tools to summarize findings, including dot plots, boxplots, and variance test results.

  • Compile key insights into a PowerPoint slide or Word document for stakeholder presentations.

This segment equips participants to identify and validate statistical differences, assess process stability, and develop data-driven recommendations for improving supplier quality, ensuring Smartboard Company’s requirements for consistent and high-performance screws are met.

In this chapter, we summarize the most important findings and the actions derived from them:

1. Sample Size Determination:

   • Conducted a hypothesis power analysis to establish the optimal sample size, ensuring reliable and statistically valid results.

2. Outlier Management:

   • Identified outliers using dot plots and excluded them by marking them with an asterisk, preventing distortion in the statistical analysis.

3. Descriptive and Graphical Insights:

   • Leveraged boxplots and descriptive statistics to understand the data's distribution, central tendencies, and variability.

4. Two-Sample T-Test Findings:

   • Confirmed a statistically significant difference: Winters screws have higher yield strength than Ziegler screws.

5. Variance Analysis Results:

   • Found Winters has greater variability in production, indicating less process control, whereas Ziegler shows consistent quality but lower yield strength.

6. Actionable Recommendations:

   • Winters: Improve process stability through quality control tools and optimization.

   • Ziegler: Enhance yield strength by refining forming and heat treatment processes.

7. Quality Assurance for Smartboard Company:

   • Provided actionable strategies to ensure consistent, high-quality screws from both suppliers, meeting Smartboard’s customer expectations.

8. Reporting with the Layout Tool:

   • Utilized the Layout Tool to create concise, professional summaries for presenting statistical findings effectively.

9. Project Preservation:

   • Saved the analysis as "2 Sample T-Test" to ensure accessibility for future use and prevent data loss.

This chapter encapsulates the critical steps and insights from the session, empowering participants to apply these analytical techniques in real-world quality management and decision-making scenarios.