Correlations in time courses of scalp electroencephalogram(EEG) may be represented by the phase synchronization in cerebral cortex sources to a certain some degree. Therefore, it is very important to localize the sources of phase synchronization and find corresponding time courses in the brain imaging study. Based on coupled Rössler oscillators with different coupling strengths, we propose a new method of simulating phase synchronized dipole sources and use a concentric 4-sphere head model to obtain simulation forward scalp EEG data. In addition, we propose the spatiotemporal dynamic analysis of phase synchronized sources based on the maximum likelihood factor analysis, verify the simulated and real scalp EEG data, and further compare the results with those of principal component analysis. Simulation results demonstrate that time courses estimated by maximum likelihood factor analysis have higher correlation with simulated sources, and less locational error between estimated sources and simulated sources. Factor analysis shows a better robust to the spatial resolution and the noise than principal component analysis. Furthermore, real data from spatial attention experiments show that factor analysis is capable of obtaining time courses and spatial distribution under the physiological base.