Path analysis (statistics)

The term path analysis referred to in the statistics a form of investigation of the dependencies between variables. As part of the path analysis are path models, ie theoretically derived models of causal relationships between variables, empirically verified. The path analysis is part of the causal analysis.

Before we proceed to model the relevant causal factors need to be identified. This problem can be solved in different ways.

The correlation between two variables is per se no causal relationship is, but delivers only the " explanandum " for a scientific explanation (see Hempel - Oppenheim scheme). In other words, the statistical correlation relationship itself provides no explanation, but must be explained theoretically self.

For example, " divorce rates " are of the "time" dependent. The independent variable "time" must be interpreted sociologically yet, so the explanation will fully delivered, as: that has changed over time, the degree of urbanization.

Demarcation

The path analysis is considered as a multiple, based on causal relationships regression analysis, however, can be considered as a special case of a structural equation model in which only individual indicators for the respective variables in the causal model can be used. This is therefore a structural equation model, in which, although the structural model is available, but should exclude the measurement model.

History

The path analysis, the geneticist Sewall Wright developed in 1918 and described in detail in 1920. Outside of genetics path analysis is used especially in sociology and econometrics.

The original claim of this statistical method to identify causal relationships has been discussed intensively from the beginning, but is now somewhat faded into the background.