Prediction intervals in the beta autoregressive moving average model

In this paper, we propose five prediction intervals for the beta autoregressive moving average model. This model is suitable modeling and forecasting variables that assume values in interval $(0,1)$. Two of proposed are based on approximations considering normal distribution quantile function distribution. We also consider bootstrap-based intervals, namely: (i) bootstrap errors (BPE) interval; (ii) bias-corrected acceleration (BCa) (iii) percentile quantiles bootstrap-predicted two different bootstrapping schemes. The were evaluated according to Monte Carlo simulations. BCa offered best performance among showing lower coverage rate distortion small length. applied our methodology predicting water level Cantareira supply system S\~ao Paulo, Brazil.