Multiverse analyses can be used to evaluate cross-lagged panel network models: An example with psychological...

Area of Science:

• Psychological science
• Quantitative psychology
• Network analysis

Background:

• Cross-lagged panel network (CLPN) models are used to estimate prospective effects between variables.
• However, CLPN models may yield spurious effects due to residual correlations and regression toward the mean.
• Existing methods often ignore model uncertainty, potentially compromising the validity of findings.

Purpose of the Study:

• To introduce and advocate for multiverse analyses as a method to enhance the rigor and transparency of CLPN models.
• To address the issue of spurious cross-lagged effects by incorporating model uncertainty.
• To provide a more reliable estimation of prospective effects in psychological research.

Main Methods:

• Recommending multiverse analyses where cross-lagged effects are estimated using various alternative models.
• Suggesting meta-analytic averaging of effect estimates from these alternative models for robust conclusions.
• Applying this methodology to re-examine cross-lagged effects between psychological flexibility and inflexibility indicators.

Main Results:

• Most previously reported cross-lagged effects between psychological flexibility and inflexibility indicators did not persist under multiverse analysis.
• The application demonstrated that alternative model specifications can significantly alter the conclusions drawn from CLPN analyses.
• This highlights the sensitivity of cross-lagged effects to model choices.

Conclusions:

• Multiverse analyses significantly improve the reliability and transparency of estimating prospective effects compared to traditional CLPN models.
• Researchers should adopt multiverse approaches to acknowledge and manage model uncertainty.
• The findings underscore the need for cautious interpretation of cross-lagged effects, particularly in the context of psychological flexibility and inflexibility.