The first step in analysing accelerated reliability testing data is to select an appropriate probability distribution. Experience shows that most of the test data collected at each stress levels can usually be appropriately fitted onto one of the underlying distributions, namely exponential, Weibull and lognormal distributions. The next step in statistical analysis is to choose an acceleration model (Arrhenius, Eyring, Inverse power law (IPL), Coffin-Manson model, etc.) that quantify the way in which probability distribution varies across different stress levels in order to extrapolate probability distribution at the level of use stress. This extrapolated probability distribution can then be used to estimate product reliability such as the probability of failure, failure rate, mean operating time between failures (MTBF) or mean time to failure (MTTF). The paper deals with a statistical analysis of failure data from accelerated reliability testing of light emitting diodes (LEDs) in military vehicles.