Related Experiment Videos
Correlates of reactivity: problems with regression based methods
1Department of Psychology, Lakehead Univesity, Thunder Bay, Ont., Canada.
Summary
Computer simulations reveal that regression methods for analyzing change scores can lead to errors, especially when a third variable correlates with baseline measurements. Change score methods are more reliable for studying reactivity relationships.
Area of Science:
- Psychology
- Physiology
- Biostatistics
Background:
- Accurately measuring reactivity is crucial in research, particularly in fields like exercise physiology and clinical hypertension studies.
- Change score and regression methods are commonly used to analyze relationships between variables and reactivity, but their statistical properties are not always well understood.
Purpose of the Study:
- To compare the effectiveness of change score and regression methods in identifying relationships between a third variable and reactivity.
- To investigate the impact of baseline correlations on the accuracy and power of these statistical methods.
Main Methods:
- Computer simulations were employed to model scenarios with and without correlations between a third variable and baseline measurements.
- The performance of change score and regression methods was evaluated based on their ability to detect true relationships and their susceptibility to Type I errors.
Main Results:
- When a third variable was not correlated with baseline, both methods performed equally well in detecting relationships with change.
- Regression methods demonstrated a high rate of Type I errors when a correlation existed between the third variable and baseline.
- Regression methods showed reduced power in detecting relationships when the direction of the effect was opposite to the baseline correlation.
Conclusions:
- The choice of statistical method is critical when analyzing reactivity, especially when baseline correlations are present.
- Regression-based change measures may inflate findings, leading to erroneous conclusions in studies examining fitness and heart rate reactivity, or hypertension and blood pressure reactivity.