<p>Prediction is a statement about a future event that can assist in making plans about possible developments. In statistics, predictive inference is used to provide an interval or tail probability of a future occurrence of a random variable based on an observed sample. Hence, predictive inference is of special interest in areas such as finance, actuarial science, medical statistics, and operational research. In this paper, a new predictive density is derived using the significance functions of the parameters obtained from the observed sample. In particular, the proposed methodology is applied to the case that the observed sample is from the inverse Gaussian distribution. A real-life example is used to show the difference in predictive inference obtained from the existing methods and the proposed method. Simulation results demonstrate the accuracy of the proposed method.</p>