In all cases, the best model chosen by all four criterion for the one variable model is with age and for the two variable model including age and sex.
The best overall model chosen by \(R^2\), adjusted-\(R^2\), and BIC is the two variable model with age and sex, but the two variable model does not improve over the single variable model by much. Mallow’s \(C_p\) chooses the two variable model over the single variable model with \(C_p=5.9\) which is close to 3. We may be missing an important predictor.