Problem 2 Flashcards
Signal detection theory
Is a method that measures the ability to differentiate between actual information and random patterns (noise) which distracts from the information
–> a number of factors affect how a detecting system will detect a signal
Psychophysical methods
Are used to determine an observers absolute threshold
BUT: however the absolute threshold may be specified differently among participants
SDT provides a precise language and graphic notation for analyzing decision making in the presence of uncertainty.
Elaborate
A 2x2 Matrix is provided, generated by 4 events that might happen when the participant responds
- Hit
- Miss
- -> not hitting - False alarm
- -> identifying when not presented - Correct rejection
Signal
Is the stimulus presented to a subject
Noise
Refers to all the other stimuli in the environment why may be mistaken for the signal
Why is noise always involved in experiments ?
Because it is always present
–> thus they either present S+N or just N
Which noise factors may affect a subjects performance ?
External or internal noise, defined by noisy neural responses
How do neurons generate a specific response in the participants?
External stimuli generate neural activity in the brain
–> there will be a higher rate of neural firing when a signal is present than when its absent
THUS: when a critical threshold is exceeded, the operator will decide yes
How do errors in responding occur on a neuronal level?
The firing rate of neurons varies due to
- random variations in the environment
- the operators brain level of neural firing
THUS: the smaller the difference in intensity between signals and noise, the greater these error probabilities become
Probability distribution
Tell us what the chances are that a given single occurs due to
- Noise (N)
- Signal and noise ( S+N)
–> an overlab in these distributions means it is difficult to know
There are 2 main components to the decision-making process of participants.
Name them.
- Subjects sensitivity to the signal
2. Subjects response criterion
Response criterion
Is determined by the rule or strategy that a subject uses
ex.: whether they focus on getting more hits or avoiding false alarms
Liberal responder
to liberal criterion
- High amount of hits
- High amount of false alarms
–> probability of both hits and false alarms is high
Neutral responder
to neutral criterion
- High amount of hits
- Average amount of false alarms
–> will rarely indicate “yes” in response to noise but very likely to signal
Conservative responder
to conservative criterion
- Average amount of hits
- Low amount of false alarms
–> never indicate “yes” to noise and rarely “yes” to signal
Beta
Refers to the ratio of activity produced by signal and noise at Xc (threshold)
–> it defines the response bias/response criterion
Sometimes circumstances dictate which strategy is best to use in responding.
Why ?
Your behavior may be determined by placing a decision criterion
ex.: when examining an ill patient it is better to be biased towards “yes” than when examining a healthy patient
How do beta and Xc interact ?
If Xc is shifted to the right, beta value will be higher than one
–> thus fewer “yes responses, hits and false alarms and vice versa
Optimal performance
Occurs when Xc is placed at the intersection of 2 curves
–> when beta=1
Payoffs
May cause the participants to change their strategy/response criterion
–> does not change the sensitivity of a subject
A subjects Sensitivity to a stimulus is determined by … ?
… the distance between the peaks of noise (N) and signal + noise (S+N) distributions along the X-axis
–> this distance affects the shape of the ROC curve
Discriminatory index (d’)
Refers to the distance between the peaks of Noise and Singal + Noise
–> Characterizes the detectability of the signal if we make the IID assumption
–> doesn’t depend on the subjects response criterion but is a true measure of internal response
AND: d means distance or separation/spread
IID Assumption
Assumes that noise follows a normal distribution with fixed variance, independent of signal strength
How can d prime be determined?
By comparing the experimentally determined ROC curve to standard ROC curves
or
from calculating the proportions of hits and false alarms in the experiment
ROC curve
- Is used to predict how each criterion will affect the subject’s hits and false alarms.
- Captures the various alternatives that are available to the subject as they move their criterion to higher + lower levels
–> false alarm rate= X; hit rate = Y
Optimal beta
Depends on the subject’s decision criterion Xc which can be influenced by payoffs
–> “optimal” = maximizing gains or minimizing losses
Signal strength
Affects the probability density functions
–> when the signal is stronger, there is less overlap between the two probability density curves. Then, the subject’s choices are not so difficult as before.
