The grass is not always greener in the neighbor's yard: Bayesian and frequentist inference methods for network autocorrelated data