Analysis of Masked Data with non-ignorable missing mechanism

In this article we consider series system products with independent lifetimes for each component in which causes of system failures might be masked Due to cost considerations or environmental restrictions. As a new approach we have defined a missing indicator for the true cause of failure which might be observed or masked. The likelihood with missing not at random (MNAR) mechanism is derived and compared with the usual likelihood for masked data via some simulation studies. The results show superior performance of our approach when non-ignorable missing mechanism is occurred