Fig. 3
Final path model. This path model, performed with structural equation modelling in LISREL 8.8, presents total path coefficients (both direct and indirect associations) between latent variables (with the hypothesised result outlined in Fig. 1.) First-order latent variables are presented in squares and the second-order latent variable, which was interpreted as a vulnerability to unhealthy behaviours, in the oval circle. Note that the study is cross-sectional and that the direction of causality in the model therefore is unknown. Model fit: χ 2 208.31 with df 50, RMSEA 0.08, GFI 0.94, AGFI 0.90 and SRMR 0.08. All path coefficients in the model were statistically significant at 95% confidence intervals except for the path between ‘age’ and ‘vulnerability to unhealthy behaviours’. * The path coefficient between ‘smoking’ and the second-order latent variable was fixed to 1.00 to standardise the second-order latent variable. † Measures high levels versus low levels