Genome-wide expression and protein profiles provide powerful tools for large-scale analyses of gene interaction and identification of pathways underlying cells' response to perturbations. Clustering algorithms, which identify distinct patterns by grouping genes with similar expression profiles, are the most widely used tools for gene expression data analysis. Although valuable, cluster analyses do not provide a complete picture of cellular processes, and more elaborate statistical and computational methods for determining these pathways (or genetic networks) must be developed. I propose a mathematical framework to describe the causal or logical relationships between gene expressions that exist in such pathways. Structural equation models are a powerful generalization of earlier statistical approaches, such as path analysis, and a widely used tool for causal inference. I employ structural equations to model relationships among genes using gene expression profiles. Solutions to the structural equations identify the pathway underlying a given causal structure or the logical relationship among the genes in the pathway. I use the method of generalized least squares to estimate the parameters in the structural equation models. Structural equation models can also assist in quantitative analysis of pathways. I have applied the proposed structural equation models to analyses of the yeast cell cycle and colon cancer apoptosis.