Target tracking: Unknown state of aircraft in a kalman filter. Factor analysis: Unknown factors to explain variation in data. Say that some of the As are missing… What happens? Effect 1: Correlation between parameters. A is binary, {0,1}. Effect 2: Multimodal posteriors. Q: What is the likelihood function when data is missing? The probability that a value of X is missing is independent of its actual value given other observed data. What if MAR does not hold? Establish an observable variable that models the probability that the data is observed given the data. If these models are binary, how many independent parameters? Why are Latent Variables a good idea?