Tables of confidence bounds for failure probabilities
The characteristics of a population cannot be determined precisely from the characteristics of a random sample taken from that population. Therefore sample characteristics can only lead to bounds for the population characteristics. The probability that these bounds are valid is called confidence. The problems of determining populatipon failure level bounds from sample failure levels and how confidence is influenced by sample size are discussed. Two sets of failure probability confidence bounds are tabulated. One set is derived for normally distributed populations and is based on the noncentral t distribution. The other set is derived nonparametrically - that is, without assuming the population distribution - and is based on the binomial distribution. These tables convert sample failure levels into population failure probability bounds. The tables are designed to be easily used and consequently are not set in the most compact format. In addition to presenting the tables, the basic statistical concepts are described, the mathematics on which the tables are based is developed, and the two methods are graphically compared. An example problem is given to illustrate the use of the tables. (auth)
- Research Organization:
- California Univ., Livermore (USA). Lawrence Livermore Lab.
- DOE Contract Number:
- W-7405-ENG-48
- OSTI ID:
- 7348587
- Report Number(s):
- UCRL-51990
- Country of Publication:
- United States
- Language:
- English
Similar Records
UPPER CONFIDENCE LIMITS FOR THE FAILURE PROBABILITY OF COMPLEX NETWORKS
Lower confidence bounds on system reliability using component data: Maximus methodology