FIG. 2.3 Results of the path analysis part of SEM models 1–4 show how the relationship between gender and physics identity is mediated through self-efficacy, interest, and perceived recognition (P.R.). (a) In model 1, there are only covariances between each pair of constructs: self-efficacy, P.R., and interest. (b) In model 2, self-efficacy predicts interest and P.R., and interest predicts P.R. (c) In model 3, interest predicts self-efficacy and P.R., and self-efficacy predicts P.R. (d) In model 4, P.R. predicts interest and self-efficacy, and self-efficacy predicts interest. The dashed lines represent residual covariances between constructs. The solid lines represent regression paths, and the numbers on the lines are standardized regression coefficients (β values), which represent the strength of the regression relations. Each regression line thickness qualitatively corresponds to the magnitude of the β value. Regression coefficients with p < 0.050 are indicated by superscript ***, p < 0.010 are indicated by superscript **, and p ≤ 0.001 are indicated by no superscript. For clarity, we have removed the statistically insignificant regression paths from gender to identity or gender to P.R. and self-efficacy in models 2 and 3. More about this image found in Results of the path analysis part of SEM models 1–4 show how the relationsh...