Kleijn, R.H.

model parameter cycle density bayesian forecast distribution statistic component weight result value hypothesis scheme combination series analysis journal exchange rates assumption likelihood exchange jeffrey alternative hypothesis figure probability score erasmus university rotterdam approach function period example forecasting vector frequency impulse response function half-life uncertainty return bayesian score statistic model mi test statistics section bayesian analysis model weights matrix portfolio impulse response method stock econometrics score statistic time-varying alternative trend growth model parameters van dijk specification variable sample strategy factor paper variance deviation time series gibbs sampler assumption 3 dem /usd time series models spectrum rotterdam application institute expectation kleibergen unit root error