<p>But what if its valence in your resume model is negative to begin with?</p>
<p>resume<em>hat = gamma1</em>othervar - gamma2_Stanford + u, where Stanford is a binary variable ~ {1, if attend; 0, otherwise, and gammai is marginal effect of components of resume.</p>
<p>i.e. d(resume_hat)/d(Stanford) = -gamma2</p>