(91) Beyond classical assumptions: Handling non-normality and heteroskedasticity in linear regression
Date:
Contributors: Rajh-Weber, H., Huber, S. E., & Arendasy, M.
Venue: 34th Conference of the Austro-Swiss Region (ROeS) of the International Biometric Society (IBS), Graz, Austria, September 14-18, 2024
Abstract: Researchers are frequently encouraged to use well known parametric methods for data analysis, even though their actual data often fails to satisfy the necessary assumptions. Even though a variety of alternative methods to ordinary least squares regression (OLS) exist, they are not available in all software packages or simply lack popularity and exposure. Using OLS regression with corrected standard errors or bootstrap methods might be a more viable solution in practice since they are better known, readily available and do not force researchers to familiarize themselves with a different estimator. In our simulation study we considered HC3 or HC4 standard errors, and the pairs and wild bootstrap resampling method with either bootstrap p-values, percentile confidence intervals, or BCa confidence intervals in addition to the standard OLS regression inference. None of the methods performed satisfactorily on all accounts tested in the study. However, using HC3 or HC4 standard errors, or a wild bootstrap procedure with percentile confidence intervals, could yield reliable results in many, but not all, scenarios. These diverse results highlight that a better understanding of available alternative inference methods, as well as knowing when and how to apply them in different situations can prove valuable for applied researchers.