Instruction: Describe the application of latent variable models in controlling for unobserved confounding in causal analysis.
Context: Candidates should explain how latent variable models can infer unobserved variables and their role in addressing omitted variable bias.