|
8.
Assessing Product Reliability
8.4. Reliability Data Analysis
|
|||
|
Graphical estimation methods (aided by computer line fits) are easy
and quick
Maximum likelihood methods are usually more precise - but require special software |
Two widely used general methods will be described in this section:
Recommendation On Which Method to Use
Maximum likelihood estimation (except when the failure data are very sparse - i.e., only a few failures) is a more precise and flexible method. However, with censored data, the method of maximum likelihood estimation requires special computer programs for distributions other than the exponential. This is no longer an obstacle since, in recent years, many statistical software packages have added reliability platforms that will calculate MLE's and most of these packages will estimate acceleration model parameters and give confidence bounds as well. If important business decisions are based on reliability projections made from life test data and acceleration modeling, then it pays to obtain state-of-the art MLE reliability software. Otherwise, for monitoring and tracking reliability, estimation methods based on computer-augmented graphical procedures will often suffice. |
||
