This thesis examines the impact of immigration on reports of mentally unhealthy days among respondents in the 2007 Los Angeles Health Survey. I argue that the findings in the extant literature are unstable due to unobserved heterogeneity in response. I fit 3 different regression models: the Poisson, the Negative Binomial, and the mixture of Negative Binomial models. From the test of goodness fits, the Mixture of the Negative Binomial models has a better fit than the other two traditional statistical models. A significant mixing proportion of my mixture model indicates that mixture of the Negative Binomial models is necessary. Two distinct distributions indicate that the model fits and identifies two kinds of people: distressed and non-distressed individuals. I use the finite mixture parameter estimates to calculate the posterior probability of being in the non-distressed group; meanwhile, I find evidence that race and economic status play important roles in classification but not migration-related factors, including years in US, citizenship and language ability.