In this case, use statistical analysis, given a % confidence we want to achieve of correctly

being able to distinguish two different temperatures. If the distribution is assumed to be

Gaussian, we can use a multitude of test to determine the resolution (students t-test is

easy).

Precision- is the number of distinguishable alternatives from which the result is selected.

In signals that have been A/D converted, this can be no more than 2^ # bits in the

converter.

Range = resolution * precision

Ex. A digital thermometer

Range: 0 to 99.9°C

Display is 3 digits XX.X, so assume resolution is 0.1 °C

Precision is 1000, or about 10 bits

Please note: Webster has another definition of precision (the quality of obtaining the

same output from repeated measurement of the same input). The word “precision” can be

used in both cases. In this class, I am most concerned with distinguishable alternatives.

Reproducibility- is the ability to give the same output for equal inputs over some period

of time (full range or standard deviation).

Specificity- is the ability of an instrument to respond to the desired input only. Signal-tonoise

ratio is one measure. The number of false positives in a binary system is another.

**Click here for download: http://www.mediafire.com/?31z8r4s9n26cu86**