A certain type of missile-borne accelerometer is the vulnerable spot that affects the storage life of the rocket control cabin. Considering that the statistical data of accelerometer is binomial and incomplete, we used isotonic regression, variable sample Pearson χ2 test of goodness of fit together with minimum χ2 estimation, to solve the “reversal” problem of failure frequency. The distribution parameters were given, and Pearson test of goodness of fit was made to the distribution function. The result showed that: both I maximum distribution and Weibull distribution were sui, and Weibull distribution was selected as the life distribution because the former one had no accelerating ability. In the end, the natural storage life under confidence level of 0. 90 and reliability of 0. 95 was obtained. The acceleration factor under 90 ℃ is 7. 2695. Thus the Accelerated Life Time (ALT) of certain batches of accelerometer was obtained. The study is of engineering significance for ALT plan of and budget.