Summary: We consider the problems of hypothesis testing and interval estimation for variance components in general linear mixed models. Exact tests and confidence intervals for a single variance component corresponding to random effects are developed based on the concepts of generalized \(p\)-values and generalized confidence intervals. The generalized \(p\)-values and generalized confidence intervals we have developed are functions of minimal sufficient statistics. Exact tests and confidence intervals are also established for comparing the random effects variance components in two independent general linear mixed models. Furthermore, we investigate the statistical properties of the resulting tests and confidence intervals. Some simulation results to compare the powers of the proposed tests with those of existing methods are reported. The simulation results indicate that the power of the new tests is a significant improvement. Finally, the proposed methods are demonstrated by a real example.