How we are misinterpreting physical activity intention - behavior relations and what to do about it

Area of Science:

• Behavioral Science
• Health Psychology
• Exercise Science

Background:

• Physical activity promotion efforts are informed by studies on the intention-behavior gap and its moderators.
• Traditional linear modeling in these studies assumes linearity and homoscedasticity, which may not always hold true.
• Factors like habit and perceived behavioral control are often examined as moderators of the intention-behavior relationship.

Purpose of the Study:

• To investigate the impact of statistical assumptions on the modeling of physical activity intention-behavior associations.
• To explore how non-linear relationships and unmet statistical assumptions affect the estimation of moderation effects.
• To determine if commonly observed moderating effects of habit and perceived behavioral control are artifacts of statistical assumptions.

Main Methods:

• Simulated data were used to model physical activity intention-behavior associations and moderation effects of habit.
• Three datasets were generated: (a) uncorrelated normal distributions, (b) correlated normal distributions, and (c) realistically correlated, non-normally distributed parameters.
• The intention-behavior association and moderation effects were analyzed across these different data scenarios.

Main Results:

• In normal distribution datasets (uncorrelated and correlated), linear moderation effects were consistent across the data range.
• A realistic dataset with non-normally distributed parameters and few individuals with low intention but high behavior showed a non-linear intention-behavior association.
• This non-linearity led to inflated linear moderation estimations for habit and perceived behavioral control, replicating findings from previous studies.

Conclusions:

• Identical correlation coefficients can obscure diverse intention-behavior associations and moderation effects.
• Unfounded statistical assumptions in linear models pose a risk of misinterpreting the physical activity intention-behavior gap and its moderators.
• The well-documented moderating effects of habit may be statistical byproducts of unmet model assumptions, necessitating a re-evaluation of existing findings.