Cortisol diurnal patterns, associations with depressive symptoms, and the impact of intervention in older adults: Results using modern robust methods aimed at dealing with low power due to violations of standard assumptions

• Modern robust statistical methods for analyzing psychobiological data are studied.
• Violations of standard assumptions can result in poor power using standard strategies.
• Improved methods for dealing with curvature make a practical difference.
• Highly nonsignificant results can become significant using modern methods.
• Cortisol associations with depressive symptoms and perceived health are described.

Advances in salivary bioscience enable the widespread integration of biological measures into the behavioral and social sciences. While theoretical integration has progressed, much less attention has focused on analytical strategies and tactics. The statistical literature warns that common methods for comparing groups and studying associations can have relatively poor power compared to more modern robust techniques. Here we illustrate, in secondary data analyses using the USC Well Elderly II study (n = 460, age 60–95, 66% female), that modern robust methods make a substantial difference when analyzing relations between salivary analyte and behavioral data. Analyses that deal with the diurnal pattern of cortisol and the association of the cortisol awakening response with depressive symptoms and physical well-being are reported. Non-significant results become significant when using improved methods for dealing with skewed distributions and outliers. Analytical strategies and tactics that employ modern robust methods have the potential to reduce the probability of both Type I and Type II errors in studies that compare salivary analytes between groups, across time, or examine associations with salivary analyte levels.