Because of its flexibility, event tracing is a commonly used technique for performance evaluation of parallel applications. To generate the time-stamps of events, a precise global time reference is needed so that the total event order reflected by a trace is coherent with the partial causal event order inherent to parallel executions. Different statistical methods to estimate global time, which are based on a linear model, have been proposed in the literature. Those methods are appropriate for performance evaluation because they do not perturb the analyzed applications. However, the estimation algorithms must collect some sample data from which global time is computed. The longer the sample collection process, the more precise the global time and the greater the delay before global time is available. This paper familiarizes the reader with statistical global time estimation methods by presenting two methods, which have been introduced in the literature. Then, we show how a good balance between length of sample period and global time precision can be achieved through a detailed experimental analysis of the estimation error observed on samples.