Comparing the regression slopes of independent groups

Area of Science:

• Statistics
• Biostatistics
• Econometrics

Background:

• Hypothesis testing for equality of slopes in regression analysis is crucial for comparing groups.
• Standard methods often assume normality and homoscedasticity, which may not hold in real-world data.
• Non-normality and heteroscedasticity can significantly impact the validity of hypothesis tests.

Purpose of the Study:

• To evaluate the performance of various statistical methods for testing equal slopes of regression lines between two independent groups.
• To assess the impact of non-normality and heteroscedasticity on these hypothesis tests.
• To identify reliable methods for practical application.

Main Methods:

• Simulation studies were conducted to generate data with varying degrees of non-normality and heteroscedasticity.
• Several recently proposed statistical tests, including those designed for heteroscedasticity, were applied.
• The Type I error rates and power of these tests were compared.

Main Results:

• Some recently developed methods, despite performing well in other contexts, showed poor performance under the specific conditions of non-normality and heteroscedasticity.
• The effectiveness of the tested methods varied significantly depending on the data characteristics.
• Two specific methods demonstrated robust performance across the simulated scenarios.

Conclusions:

• The presence of non-normality and heteroscedasticity poses challenges for testing the equality of regression slopes.
• Existing methods that account for heteroscedasticity are not universally effective in this specific hypothesis testing scenario.
• Two recommended methods offer reliable solutions for comparing regression slopes in the presence of non-normality and heteroscedasticity.