Latest developments in Generalized Inference can tackle problems that once thought were not possible without resorting to asymptotic approximations or without Bayesian approach. They range from obtaining superior estimates of BLUPs in Mixed Models that do not suffer from the problem of equal BLUPs frequently given by MLE based methods to obtaining estimates of parameters of known sign.

These methods do not require one to treat parameters as random variables or deal with prior distributions and hyper-parameters. Yet in constructing confidence intervals and tests in applications involving non-normal distributions, they allow us to write exact probability statements that are valid for any sample size. Furthermore, in applications dealing with parameters of known signs (e.g. variance components are supposed to be positive) or expected sign (e.g. price elasticity of demand is expected to be negative) or range (e.g. variance ratios in BLUP formulas are supposed to be between 0 and 1), generalized inference allow one to take advantage of such knowledge and produce estimates and confidence intervals that naturally fall in the expected range.