The dynamical character of education and the complexity of its constituent relationships have long been recognized, but the full appreciation of the implications of these insights for educational research is recent. Most educational research to this day tends to focus on outcomes rather than process, and rely on conventional cross-sectional designs and statistical inference methods that do not capture this complexity. This presentation focuses on two related aspects not well accommodated by conventional models, namely fractality (self-similarity, scale invariance) and power law distributions (an inverse relationship between frequency of occurrence and strength of response). Examples are presented for both phenomena based on my empirical work on of daily high school attendance rates over time. We will discuss how the statistical indicators are generated and interpreted and what they reveal about the underlying dynamics of school attendance behavior.