When maintainability-related data is to be used for compliance testing and for determination testing, the analysis procedures used need to be considered very carefully and discussed in detail in any subsequent test report.
Matters of importance include:
- Data Editing/Data Transposition
- Distribution Analysis
- Parameter Computation
- Presentation of Results
a. Data Editing/Data Transposition - Actions taken to assure the accuracy, completeness and validity of the data should be described. If any censoring is performed, the rules and reasons for performing the censoring should be presented. If data are transposed from one form to another (e.g., from a linear to a logarithmic scale), the reason and justification for such a transposition should be clearly stated.
b. Statistical Distribution Analysis - If the data are to be analyzed statistically, it is usually
necessary to determine the underlying distribution. The most commonly used
distribution functions in maintainability are: the log-normal distribution and
the multimodal distribution (in special cases only). The method of testing the
distribution assumption should be described, with the reasons for that
specific selection. Common methods used in maintainability analysis include
the 2 (chi- square), Kolmogorov-Smirnov and various graphical tests. The Kolmogorov-Smirnov (also known as d-test) is the most frequently used method for distribution testing. Its usage, relative to maintainability data, is described in detail in IEC 706-6, "Guide on Maintainability of Equipment - Part 6: Section 9: Statistical Methods in Maintainability Evaluation," Annex C, "Kolmogorov-Smirnov Distribution Testing." For other possible statistical analysis methods pertinent statistical textbooks should be consulted.
c. Parameter Computation - The basis for computing all maintainability parameters to be presented should be clearly stated. If selected parameters are to be computed on a cumulative or interval basis, the method to be used should be detailed. If maintainability mathematical models are to be used, they should be fully described.
d. Presentation of Results - When the results are to be presented, all conditions needed for their understanding and use should be clearly stated. These conditions include the purpose of the data collection scheme, especially with respect to type and variation of the data chosen. Circumstantial information should also be provided, such as time (e.g., busy hours), locations (e.g., geographic conditions) and the current duration of the data-collection scheme. Particular situations which may limit the data application and use should be indicated (for example, any difficulties encountered, assumptions, or incompleteness of data).
Consideration should also be given to the form of presentation. A condensed form (for example, diagrams, histograms, graphical presentations) may be more appropriate than detailed numerical listings.