Precision Requirements
A
precision requirement is a
correctness
requirement that specifies a required amount of precision,
which is a
quality factor that is defined as follows:
- Precision
- the dispersion of quantitative data.
The typical objectives of a precision requirement are
to:
- Ensure that the resolution of numerical data is maximized
(i.e., that its dispersion is minimized).
Precision requirements are typically specified in terms of
the following measurements:
- Confidence intervals.
- Standard deviations.
The following are typical examples of precision
requirements:
- “The 95% confidence interval for the values of the
temperatures of the chemical mixing tanks shall be no more
than 1° Celsius.”
- “The values of the temperatures of the chemical
mixing tanks shall no vary more than 1.25° Celsius from each
other.”
The following guidelines have been found to be useful when
producing precision requirements:
- The scope of a precision requirement can be:
- Precision requirements can be identified and specified in
term of the following:
| Component of
Requirement |
Possibile Values |
| Numerical Variable |
TBD |
| Precision of the Variable |
TBD |
- Numerical data can be accurate without being precise, and
precise without being accurate.
- Precision is typically measure in terms of the maximum
permitted variability, standard deviation, of confidence
interval of quantitative data from its average value.
- For example, a temperature sensor is precise to .1 degree
Celsius, but not very accurate because it consistently under
measures the temperature by 3.5 degrees. Thus, the standard
deviation of the temperature measurements may be sufficiently
precise, but nevertheless useless because the average
temperature measurement returned by the temperature sensor is
3.5 degrees less than the true temperature making the
temperature sensor not adequately accurate.