First page
Back Continue Last page Summary GraphicsOnce you have replicated an experiment, the noise degrading sources can be modeled as sources unique to that particular presentation of the stimulus and sources common to all presentations of that stimulus.
This model of noise was proposed by former students of the department, Alan Taylor and Rick Boven.
Unique noise is a statistical concept. It is the inconsistency of an observer in responding to repeated presentations of the same stimulus. That is, the observer does not necessarily respond "Yes" the signal was present every time the same stimulus is presented.
Possible sources of unique noise are internal noises such as heartbeat, and random neural firings, as well as external noises such as a Gaussian noise background masker, a door banging, or a car going past.
The common noise is defined negatively as the complement of the unique noise. Common noise ideally results from the acoustical waveforms or stimulus set which are constant over replications of an experiment. It is what the experimenter is interested in measuring.
Unfortunately, common noise can also come from time order and sequential effects, and events correlated with elements in the stimulus set (say perhaps a 50 Hz hum from the mains).