The　analysis　of　panel　data　is　frequently　complicated　by　missing　data　due　to　information　dropout.　We　assume　that　responses　follow　a　semiparametric　varying　coefficient　random　effect　model.　The　concept　of　two　step　estimation　is　used　in　order　to　provide　consistent　estimators　of　the　parametric　and　nonparametric　parts.　The　methods　simplify　the　calculations　and　avoid　the　collinearity　caused　by　more　parameters　in　the　model.