Microvariability consists in small time scale variations of low amplitude in the photometric light curves of quasars, and represents an important tool to investigate their inner core. Detection of quasar microvariations is challenging for their non-periodicity, as well as the need for high monitoring frequency and high signal-to-noise ratio. Statistical tests developed for the analysis of quasar differential light curves usually show either low power or low reliability, or both. In this paper we compare two statistical procedures that include several stars to perform tests with enhanced power and high reliability. We perform light curve simulations of variable quasars and non-variable stars, and analyze them with statistical procedures developed from the F-test and the analysis of variance. The results show a large improvement in the power of both statistical probes, and a larger reliability, when several stars are included in the analysis. The results from the simulations agree with those obtained from observations of real quasars. The high power and high reliability of the tests discussed in this paper improve the results that can be obtained from short and long time scale variability studies. These techniques are not limited to quasar variability; on the contrary, they can be easily implemented to other sources such as variable stars. Their applications to future research and to the analysis of large field photometric monitoring archives can reveal new variable sources.