A　local　quasi-likelihood　estimator　for　generalized　semi-parametric　models　when　the　responses　are　missing　at　random　is　proposed.　It　is　shown　that　the　asymptotic　biases　are　the　same　as　when　no　data　are　missing,　while　the　variances　are　generally　different.　Furthermore,　we　demonstrate　that　the　estimator　has　asymptotically　normal　properties.