LogLogDiagramm

LogLogDiagramm, also known as a log-log diagram or double logarithmic plot, is a graphical method used to analyze relationships between two variables that span multiple orders of magnitude. In this diagram both axes are scaled logarithmically, enabling examination of multiplicative and scaling patterns. In German-language literature it is often referred to as LogLogDiagramm or Log-Log-Diagramm.

If the underlying relation is y = a x^b with a > 0 and x > 0, taking logarithms gives log y = log a + b log x. On a log-log plot the data form a straight line with slope b and intercept log a. The logarithm base is arbitrary; only the slope is invariant. Values must be positive, since the logarithm of zero or negatives is undefined.

Usage and interpretation: A straight line indicates a power-law relation; the slope estimates the exponent b, while the intercept gives the scale factor a. It is widely used in physics, biology, network science, and economics to identify scaling laws. Regression on log-transformed data provides parameter estimates, but care is needed regarding bias and heteroscedasticity, as errors may not be evenly distributed in log space.

Limitations: Not all relationships are power laws; deviations from linearity point to alternative models. Zero or negative values are problematic and require preprocessing or alternative methods. Measurement errors can influence fit differently in log space, so robust or orthogonal regression approaches may be appropriate in some cases. The LogLogDiagramm remains a useful exploratory tool for revealing scaling and multiplicative structures in data.