Signal Detection Theory and other Stats Stuff Flashcards
What is the difference between external noise and internal noise?
External noise is noise in the environment
Internal noise is noise (or bias) in the individual
What does a Probability of Occurrence graph tell us?
It lets us see the probability of a response (hit, miss, False alarm, or correct rejection) based on the level of the stimulus intensity.
What is a criterion line?
A criterion line is the point of intensity of the stimulus at which the person changes from a “no” to a “yes” in whether they detect it (regardless of whether they are correct or not.
What do Probability of occurrence graphs and criterion lines combine to tell us?
When a criterion line is applied to a probability of occurrence graph, we can determine the ratio of hits, misses, false alarms, and correct rejections.
What does it mean when the curves of a Probability of occurrence graph overlap? What about when they do not overlap as much?
The level of overlap tells you what the relationship between hits and false alarms/ misses and correct rejections is going to be. If there is a great deal of overlap it means that as the hits go up, so will the false alarms (and likewise for misses/CR). Less overlap means that the you will not go up as significantly in one just because the other increases.
What do ROC curves tell you? How is d’ related to this?
ROC crves capture various alternatives available to the individual as the criterion changes. In an ROC curve, hits are on the Y axis and False alarms are on the X axis. When a person’s sensitivity is higher (they are better able to distinguish between the stimulus and noise, the curves are further apart on the POC graphs), the ROC curve is MORE bowed.
d’ tells you how bowed an ROC curve is going to be and and how far apart the curves are on the POC graph. The GREATER the d’, the MORE bowed the curve is, the FURTHER apart the curves are on the POC graph and BETTER we can distinguish between the noise and the stimuli.
What are two ways you can reduce the amount of overlap between noise and noise + stimuli curves on POC graphs?
- Increase the amount of information available (which makes the signal more distinct) –> this moves the curves further away from each other.
- Reduce the amount of noise. This makes the curves more narrow, reducing their spread.
