The precise form of statistical analysis of data is specific to each use and can be a complex and time-consuming process. It should be carried out by an experienced analyst who can properly assess the information required to be extracted from the raw data.

Data (including maintenance data) are frequently analyzed to obtain statistical inferences regarding a given population of data. Statistical inference, is the process of drawing conclusions about an entire population of similar objects, events, or tasks, based upon a sample of a few.

Two basic approaches to statistical inference are mainly used^{17} (either or both approaches may be used in the analysis of maintenance/maintainability data^{18}):

*Parametric* -which is primarily concerned with inference about certain summary measures of distributions (mean, variance, etc.). This approach is based on explicit assumptions about the normality of population distributions and parameters.
*Distribution* -which is concerned with inference about an entire probability distribution, free of the assumptions regarding the parameters ofvthe population sampled.

Meaningful data handling and its subsequent evaluation also require some prior investigation of the process generating the data. Different sets of data available on an item may be combined, provided that the same selection criteria have been applied to each set. The choice of appropriate methods of data evaluation may be influenced by such factors as possible time-dependency of the process or more than one cause relating directly to the data.

Any peculiarities in the data collection scheme should be taken into account in developing the data and in the analytical process. The analyst should identify any data falling outside a pre-set range. Acceptance or rejection criteria should be explicitly validated.

Frequently one of a number of types of statistical distribution will underlie the collected data. Three principal methods are available to identify a particular underlying distribution:

- Engineering judgment, based upon an analysis of the physical process generating the data
- Graphical methods using special charts, leading to the construction of nomographs
- Statistical tests, such as the Chi-square and goodness of fit, providing a measure of the deviations between the sample and the assumed distributions

^{17}Hays,
W. L. and Winkler, W. L. "Statistics-Probability, Inference and Decision",
Holt, Reinhart and Winston, New York, 1971

^{18}Knezevic, J. "Effective Analysis of Existing
Maintainability Data", SAE Communications in RMS, Volume 2/Number 1, January
1995