Supplementary table. Mixed effect meta-analysis: methodological characteristics as potential moderators of habit-behavior effects (N = 6173†; k = 23)

Meta-analysis of habit mean scores†† / Meta-analysis of habit-behavior correlations
N / k / QB / Mean (%)
(95% CI) / QW / I2 / QB / r+
(95% CI) / QW / I2
% / On 1-7 scale
Study design / 0.18 / 0.00
Cross-sectional / 9 / 2942 / 48.10***
(41.67, 54.54) / 3.89***
(3.50, 4.27) / 684.32*** / 98.83 / .43***
(.40, .46) / 61.30*** / 86.95
Prospective / 14 / 3231 / 49.87***
(44.68, 55.06) / 3.99***
(3.68, 4.30) / 313.39*** / 95.85 / .46***
(.43, .49) / 159.44*** / 91.85
Sample / 0.82 / 0.38
University-based / 14 / 2500 / 49.53***
(44.87, 54.18) / 3.97***
(3.69, 4.25) / 357.02*** / 96.36 / .48***
(.45, .51) / 119.28*** / 89.10
School students / 5 / 1196 / 50.94***
(43.22, 58.67) / 4.06***
(3.59, 4.52) / 352.96*** / 98.87 / .42***
(.37, .46) / 9.43 / 57.59
General adult population / 4 / 2477 / 45.80***
(37.23, 54.37) / 3.75***
(3.23, 4.26) / 17.76*** / 83.11 / .42***
(.39, .46) / 87.47*** / 96.57
Behavior measure
Self-report vs objective / - / 2.31
Self-report / 21 / 5544 / - / - / - / - / .46***
(.44, .48) / 197.03*** / 89.85
Objective / 2 / 629 / - / - / - / - / .28***
(.20, .35) / 0.15 / 0.00
Actual vs typical / - / 1.44
Actual / 13 / 2950 / - / - / - / - / .51***
(.48, .53) / 142.27*** / 91.57
Typical / 10 / 3223 / - / - / - / - / .37***
(.34, .40) / 35.86*** / 74.90
Study quality††† / - / 0.71
Low quality / 7 / 1660 / - / - / - / - / .52***
(.48, .55) / 4.50 / 0.00
Acceptable quality / 16 / 4513 / - / - / - / - / .42***
(.39, .44) / 200.12*** / 92.51

*** p≤.001. † N is less than sum of study Ns from Table 2 because sample sizes have been reduced in two instances, in each of which a single dataset generated two habit-behavior correlations. †† Behavior measures, and so study quality, were not expected to influence habit strength and hence were explored as potential moderators of habit-behavior correlations only. †††Studies were treated as low quality where they used a cross-sectional design and were based on self-reported typical behavior. All other studies were deemed of acceptable quality. QB: between-subgroup heterogeneity attributable to the moderator. QW: within-subgroup heterogeneity.