Intro to Business Statistics: Analysis of Variance ANOVA TwoWay ANOVA Main Effects Interaction Effects

What This Is

Two-way ANOVA (Analysis of Variance) is a statistical method used to analyze the effects of two independent variables on a continuous dependent variable. This technique helps businesses understand how the interaction between two factors affects the outcome, enabling informed decisions. For instance, a retail chain wants to know if average daily sales exceed $10,000 when considering both the type of product (e.g., electronics, clothing) and the store location (e.g., urban, suburban).

Key Formulas & Symbols

• F = (MSB / MSE) where MSB = Mean Square Between, MSE = Mean Square Error, and F is the F-statistic.
• MSB = Σ(n_ij * (x̄ij - x̄...))² / (a - 1) where n_ij = sample size for group i in factor j, x̄ij = sample mean for group i in factor j, x̄... = grand mean, and a = number of levels in factor A.
• MSE = Σ(n_ij * (x_ij - x̄_ij)²) / (N - a - b + 1) where N = total sample size, x_ij = individual data point for group i in factor j, and b = number of levels in factor B.
• SSA = Σ(n_ij * (x̄i... - x̄...))² where SSA = Sum of Squares for factor A, n_ij = sample size for group i in factor j, x̄i... = mean for factor A level i, and x̄... = grand mean.
• SSB = Σ(n_ij * (x̄...j - x̄...))² where SSB = Sum of Squares for factor B, n_ij = sample size for group i in factor j, x̄...j = mean for factor B level j, and x̄... = grand mean.
• SSE = Σ(n_ij * (x_ij - x̄_ij)²) where SSE = Sum of Squares for Error.
• SSA + SSB + SSE = SST where SST = Total Sum of Squares.
• x̄_ij = (Σx_ij) / n_ij where x̄_ij = sample mean for group i in factor j, Σx_ij = sum of individual data points for group i in factor j, and n_ij = sample size for group i in factor j.
• x̄_... = (Σx_ij) / N where x̄_... = grand mean, Σx_ij = sum of all individual data points, and N = total sample size.

Step-by-Step Procedure

1. State hypotheses: Formulate null and alternative hypotheses for the main effects and interaction effect. For example, H₀: μ_A1 = μ_A2 = μ_A3 and H₀: μ_B1 = μ_B2 and H₀: μ_A1B1 = μ_A1B2 = μ_A2B1 = μ_A2B2 = μ_A3B1 = μ_A3B2 = 0, H_a: At least one μ_A ≠ μ_A' or at least one μ_B ≠ μ_B' or μ_A1B1 ≠ μ_A1B2 ≠ μ_A2B1 ≠ μ_A2B2 ≠ μ_A3B1 ≠ μ_A3B2.
2. Choose test: Select the F-test for two-way ANOVA.
3. Compute test statistic: Calculate the F-statistic using the formulas above.
4. Find p-value or critical value: Determine the p-value associated with the F-statistic or find the critical F-value from the F-distribution table.
5. Compare to α: Compare the p-value to the significance level α (default = 0.05) or compare the F-statistic to the critical F-value.
6. Conclude: Based on the comparison, reject the null hypothesis if the p-value < α or the F-statistic > critical F-value, indicating a statistically significant effect.

Common Mistakes

• Mistake: Failing to check the assumptions of two-way ANOVA (normality, equal variances, independence).
• Correction: Verify the data meets the assumptions before proceeding with the analysis.
• Mistake: Misinterpreting the interaction effect as a main effect or vice versa.
• Correction: Clearly distinguish between the interaction effect and main effects in the interpretation.
• Mistake: Failing to account for the degrees of freedom when calculating the F-statistic.
• Correction: Ensure to use the correct degrees of freedom (df_A = a - 1, df_B = b - 1, df_E = N - a - b + 1) when calculating the F-statistic.

Quick Practice Problems

1. A company wants to know if the average sales of its two products (A and B) differ across three regions (North, South, East). The data is as follows:

What is the p-value for the interaction effect?

Answer: 0.023, The p-value is calculated using the F-statistic and the F-distribution table.

1. A marketing firm wants to know if the average response to two advertising channels (TV and Radio) differs across three age groups (18-24, 25-34, 35-44). The data is as follows:

What is the F-statistic for the main effect of TV?

Answer: 4.23, The F-statistic is calculated using the formulas above.

1. A quality control team wants to know if the average defect rate of three production lines (A, B, C) differs across two shifts (Morning, Afternoon). The data is as follows:

What is the p-value for the main effect of Shift?

Answer: 0.001, The p-value is calculated using the F-statistic and the F-distribution table.

Last-Minute Cram Sheet

1. F-statistic: F = (MSB / MSE), where MSB = Mean Square Between and MSE = Mean Square Error.
2. Degrees of freedom: df_A = a - 1, df_B = b - 1, df_E = N - a - b + 1.
3. Assumptions: Normality, equal variances, independence.
4. Interaction effect: The effect of the interaction between two factors on the outcome.
5. Main effects: The effects of each factor on the outcome, independent of the other factor.
6. p-value: The probability of observing the data (or more extreme) if the null hypothesis is true.
7. α: The significance level (default = 0.05).
8. Critical F-value: The F-value that separates the rejection region from the non-rejection region.
9. ⚠️ p-value is NOT the probability that H₀ is true – it’s the probability of observing the data (or more extreme) if H₀ is true.
10. ⚠️ F-statistic is calculated using the formulas above, not just the ratio of means.