This　paper　is　to　study　the　validity　of　the　G-diagnosis　algorithm　that　has　been　regarded　as　a　classical　algorithm　for　faulty　diagnosis　in　analog　circuits.　The　results　of　this　study　indicate　that:　(1)　wrong　mathematical　models　are　used　to　estimate　G-parameters;　(2)　the　estimated　G-parameters　must　be　biased　against　actual　G-parameters;　(3)　the　G-algorithm　will　give　mistaken　diagnosis　conclusions　if　harsh　conditions　can　not　be　satisfied.　The　paper　proposes　formulas,　which　are　easily　calculated　with　nominal　data　of　circuits,　to　forecast　if　the　G-diagnosis　algorithm　can　locate　faults　correctly　or　not.