Lecture 6: Integration of results and goodness of decision

Question 1

Clinical (1) vs. statistical (2) prediction

Answer 1

1) human judgement, available information can be combined freely (no clear rules)
2) application of an a pirori set rule, which ideally has been empirically validated

Question 2

Reasons for errorenuos clinical judgements (What are the different errors?)

Answer 2

1. Strategy-based errors
2. Association-based errors
3. Psychophysical-based errors

Question 3

Strategy-based errors

Answer 3

exponses for well-designes strategy are thought to be too high - hence, people only use suboptimal strategies

Question 4

Association-based errors

Answer 4

• Info is connected with wrong or irrelevant associations
• cognitive distortions

Question 5

Psychophysical-based errors

Answer 5

erroneous assessment of costs and utility of individual information.

Question 6

Potential counter measures
(against erroneous judgements)

Answer 6

Highlight the importance of the diagnosis and the professional responsibility (of the assessor)

Question 7

What kind of prediction is better - clinical or statistical?

Answer 7

• statistical prediction is better in general
• but often only small differences
• when using interview data, statistical prediction is significantly superior (moderator)

Question 8

What are potential reasons for superiority of the statistical prediction?

Answer 8

• humans often ignore the base rate
• they tend to weigh information incorrectly
• they often ignore statistical phenomena such as regression to the mean.

Question 9

Criticism of statistical prediction despite its advantages

Answer 9

• often (unrealistic) restriction to one test only
• restriction to info that is availabe for ALL individuals
• only applicable if empirically validated computaion rules are available

Question 10

What can be done? - In order to enhance statistical predictions?

Answer 10

• development and usage of systematic scoring keys
• Time- and Event Sampling for behavior observation
• take measurement error into account

Question 11

What is time sampling?

Answer 11

An observation section is divided into equal time intervals - then it is counted whether the target behavior is shown in the sections.

Question 12

What is event sampling?

Answer 12

Here we count how often the behavior is performed in total during the observation period.

Question 13

Why is the interpretation of norm scores problematic?

Answer 13

• we need to know and acknowledge the specifics of the chosen comparison group
• measurement error is inherent in all test scores.

Question 14

Why do we compute a confidence interval?

Answer 14

because there is always (to some degree) error in test scores.

Question 15

What leads to smaller Confidence intervals?

Answer 15

a higher reliability

Question 16

What are the different choices of the reliability type?

Answer 16

For prognosis: Retest reliability
For current state: Internal consitency

Question 17

What is the difference for a directed and an undirected hypothesis?

Answer 17

• for directed hypothesis, we test against the treshold.
• for undirected hypothesis - both sides of the treshol are relevant.

Question 18

What are problems associated with the use of confidence intervals?

Answer 18

• difficult to understand for lay people
• methodological: (assumptions of symmetry, assumption that measurement error in of the same size throughout the latent variable continuum)

Question 19

What are two important factors of the test profile?

Answer 19

1. All test /scales should be reliable
2. High intercorrelations between test/scales lower profile reliability.

Question 20

What are potential reasons of discrepancies? (when integrating results)

Answer 20

• within the testee
• between assessed constructs
• level of abstraction inherent in the measure
• specific demans/ requirements posed by different methods.
• failure to ensure objectivity
• different ways of test norming.

Question 21

Sensitivity

Answer 21

• proportion of true positive tests out of all test takers within the condition (hit rate)
• ability of a test to correctly diagnose a test taker who does have the condition

Question 22

Specificity

Answer 22

• proportion of true negative tests out of all test takers without the condition
• ability of test to correctly diagnose a test taker who does NOT have the condition.

Question 23

Positive predictive value

Answer 23

probability of a test, when returning a positive result, of correctly identifying the true positives, and at the same time avoiding the all false positives.
(JUST LOOKING AT POSITIVES)

Question 24

Negative preditive value

Answer 24

probability of a test, when returning a negative result, of correctly identifying the true negatives, and at the same time avoiding all false negatives.
(JUST LOOKING AT NEGATIVES)

Question 25

Hoes does the Goldberg Index work?

Answer 25

Lie scale + Paranoai + Schizophrenia - Hysteria - Psychasthenia (T-values)
- is index is > 45, then the patient is judged to be psychotic.

Question 26

What does the measurement error tell us?

Answer 26

gives us an idea of the precision we have achieved with our measurement.