Covariance analysis, positivity and the Yakubovich-Kalman-Popov lemma