* p <.001

Mediation analyses were conducted for each anxiety disorder measure using Baron and Kenny’s (1986) method. All assumptions of multiple regression were met, with linearity between variables and no evidence of multicollinearity, apart from OCI-R scores which demonstrated evidence of substantial positive skewness and were transformed using logarithmic transformation (Log10). Additionally, SMSP scores showed evidence of moderate positive skewness and were transformed using the square-root method, as suggested by Tabachnick and Fidell (2007) and Howell (2007).

Using GAD-7 scores as the IV in Step One of the mediation model, GAD scores significantly predicted AQ scores, β =.20, SE =.09,t (358)=3.77, p <.001, accounting for 3.6% of the variance, F (1,358)=14.27, p <0.001. Step Two showed that GAD was also a significant predictor of the mediator IU (measured by IUS-12) β =.38, SE =.10 t (358)=7.86,p <.001, and accounted for 14.5% of the varianceF (1,358)=61.70, p <0.001. Step Three showed that IU was a significant predictor of AQ, β =.49, SE =.04t (358)=10.65, p <.001, predicting 23.8% of the variance, F (1,358)=113.34, p <0.001. Step Four showed that when controlling for IU, the model accounted for 23.6% of the variance F (2,357)=56.54, p <0.001, and GAD scores were no longer a significant predictor of AQ β =.01,SE =.009 t (358)=.18, p =.856, while IU remained significant, β =.49, SE =.04 t (358)=9.75,p <.001, thus IU is a partial mediator of the relationship between GAD and AQ, as demonstrated in Figure 1a.

The second part of the GAD model aimed to establish if SOR was a mediator in the relationship between GAD and AQ. Step One describing GAD as a significant predictor of AQ has been described above. Step Two demonstrated that SOR (as measured by ASPQ) could be significantly predicted by GAD, β =.30, SE =.25, t (358)=5.93,p <.001, and predicted 8.7% of the variance in SOR scores, F (1,358)=35.21, p <0.001. In Step Three, SOR was also a significant predictor of AQ scores, β =.53,SE =.02 t (358)=11.85, p <.001, predicting 28% of the variance F (1,358)=140.38, p <0.001. Step Four demonstrated that when both GAD and SOR were used to predict AQ scores, GAD was no longer significant, β =.04, SE =.04,t (358)=.87, p =.388, while SOR remained a significant predictor, β =.520, SE =.52, t (358)=11.04,p <.001, thus SOR was a partial mediator in the relationship between GAD and AQ, with both variables explaining 27.9% of the variance in AQ scoresF (2,357)=70.51, p <0.001, as illustrated in Figure 1b.