Planned Missing Data Designs for Spline Growth Models in Salivary Cortisol Research

Abstract

Salivary cortisol is often used as an index of physiological and psychological stress in exercise science and psychoneuroendocrine research. A primary concern when designing research studies examining cortisol stems from the high cost of analysis. Planned missing data designs involve intentionally omitting a random subset of observations from data collection, reducing both the cost of data collection and participant burden. These designs have the potential to result in more efficient, cost-effective analyses with minimal power loss. Using salivary cortisol data from a previous study (Hogue, Fry, Fry, & Pressman, 2013), this article examines statistical power and estimated costs of six different planned missing data designs using growth curve modeling. Results indicate that using a planned missing data design would have provided the same results at a lower cost relative to the traditional, complete data analysis of salivary cortisol.

Keywords:

• planned missing data design
• salivary cortisol
• measurement efficiency
• growth curve modeling
• missing data

ACKNOWLEDGMENTS

This study was supported by grants from the Association for Applied Sport Psychology and The University of Kansas’ School of Education Graduate Research Fund. Additional support for this project came from grant NSF 1053160 (Wei Wu & Todd D. Little, co-PIs) and by the Center for Research Methods and Data Analysis at the University of Kansas (when Todd D. Little was director). Todd D. Little is now director of the Institute for Measurement, Methodology, Analysis, and Policy at Texas Tech University. The authors would like to acknowledge Dr. Andrew C. Fry and Dr. Sarah D. Pressman for their assistance with the original research study. Author contribution by the first two authors was equivalent.

Notes

¹Although other approaches, such as power analysis on absolute model fit (MacCallum, Browne, & Sugawara, 1996) and nested model comparison (MacCallum, Browne, & Cai, 2006), do not require parameter values, the Monte Carlo approach is more general in that it can find the power of each parameter estimate in any models and find the power with PMDD (Muthén & Muthén, 2002).

²Power analysis for PMDD may involve intensive computer programming. Applied researchers may use the simsem package (Pornprasertmanit, Miller, & Schoemann, 2013) in R, which can provide power analysis using a Monte Carlo simulation with PMDD (see simsem.org for details).