Since the first experimental realization of Bose-Einstein condensates, ultracold atomic gases have been one of the most promising platforms for exploring quantum many-body physics. While zero-temperature mean-field theories successfully describe various equilibrium and dynamical phenomena, realistic experimental systems inevitably operate at finite temperatures, where thermal fluctuations always exist. Thermal fluctuations not only deplete the condensate fraction but can also lead to unexpected physical effects, e.g., temperature-driven superfluid-supersolid phase transitions in dipolar gases. In this review, we summarize three major theoretical approaches for treating finite-temperature effects in ultracold Bose gases: the temperature-dependent extended Gross-Pitaevskii equation (TeGPE), classical-field methods based on projected Gross-Pitaevskii equations (PGPE/SPGPE), and quantum Monte Carlo (QMC) techniques. Specifically, we mainly discuss their underlying physical assumptions, regimes of applicability, and representative applications to equilibrium phase diagrams, dissipative dynamics, and fluctuation-dominated phenomena. The TeGPE incorporates thermal and quantum fluctuations through effective mean-field potentials within the local density approximation, providing an efficient description of equilibrium properties and phase transitions. Classical-field approaches, including PGPE and SPGPE, describe highly occupied low-energy modes and are particularly suitable for studying thermalization, dissipative evolution, and non-equilibrium dynamics. In contrast, QMC methods, especially path integral Monte Carlo, preserve the full Bose statistics and many-body correlations, providing reliable benchmarks for strongly correlated finite-temperature systems. Finally, we outline recent developments toward quantum kinetic descriptions that permit a self-consistent description of the interplay between condensates and thermal clouds in real-time dynamics.