Week 4

Question 1

The difference between cause and coincidence is…?

Answer 1

correlation does not mean causation

Question 2

To say medication cured the headache is a ___ ____

Answer 2

causal inference

Question 3

Treatment effect =

Answer 3

amount of observed change in outcome

Question 4

What is spurious causation?

Answer 4

occurs when two factors appear casually related to one another but are not

Question 5

Direction of causality can go both ways… e.g.

Answer 5

Social media makes teenagers depressed
OR
depressed teenagers tend to use social media more

Question 6

Confounding =

Answer 6

“third-variable problem”

Question 7

Selection (sampling) bias is…?

Answer 7

• Certain types of people recruited–> cannot generalise results to population

Question 8

Allocation bias is…?

Answer 8

starting differences between treatment and control conditions –> outcome differences may be due to starting differences and not the treatment (SOLUTION: run RCT)

Question 9

Maturation bias is…?

Answer 9

• Changes occur naturally over time, not due to treatment

- SOLUTION: RCT, compare treatment and control, both should change the same

Question 10

Detection bias is…?

Answer 10

• Differences between groups in how outcomes are measured

- SOLUTION: Blinding assessors measuring outcomes

Question 11

Performance bias is…?

Answer 11

• Systematic differences in how treatment is provided

- SOLUTION: blind participants and clinicians, maintain control over procedures (it is difficult)

Question 12

Attrition bias is…?

Answer 12

• Differences between groups among those dropping out

- SOLUTION: maximise retention, conduct short term rather than long term

Question 13

Measurement bias =

Answer 13

systematic error, invalid results

Question 14

Regulation bias may be…?

Answer 14

ethical or bureaucratic restrictions on research

Question 15

Population choice biased –> some groups e.g. minors, vulnerable, rural, remote, threatens ____ validity

Answer 15

external (generalisability)

Question 16

Attrition bias threatens ____ validity

Answer 16

internal

Question 17

Selective reporting bias is..?

Answer 17

Researchers only show desired result

Question 18

Publication bias is…?

Answer 18

Only significant results reported

Question 19

Author prestige bias is….?

Answer 19

A known author and researcher being more trusted than a new one (despite their work)

Question 20

A case series design with just a single observation after treatment cannot detect ____

Answer 20

change (before and after difference)

Question 21

Can the test itself interfere with validity?

Answer 21

Yes, the before or pre-test might act as practise and make the person more competent when they take the second (false improvement)

Question 22

Statistical regression is a risk when selection for treatments occurs because of unusually ….?

Answer 22

good or poor performance is prevalent before any treatment (someone may ‘accidentally’ perform very badly, and then when they perform at their usual standard it looks like an improvement)

Question 23

Comparative studies are also known as…?

Answer 23

quasi-experiments

Question 24

Comparative study is level ___

Answer 24

III

Question 25

RCT is level ___

Answer 25

II

Question 26

Problem with quasi-experiments...

Answer 26

- Post treatment differences between groups may be attributed to treatment, when there were actually differences before hand simply in the groups - Treatment or control groups may have differences that have not been measured that help or stop treatment

Question 27

Internal validity threats that random allocation can't protect (5)

Answer 27

- Mortality/attrition bias - Statistical conclusion validity - problems with statistical accuracy - compensatory rivalry - control group tries to prove themselves - imitation or compensatory alternative treatment that control receives - resentful demoralisation - control group goes on strike because they are 'missing' treatment)

Question 28

Single blinding vs double blinding?

Answer 28

Single blinding: allocation concealed to participants Double blinding: Allocation also hidden from staff and researchers (where possible)

Question 29

External validity threats that random allocation can't protect (6)?

Answer 29

- Artificial nature of many experiments (lab setting) - Treatment inference (in real world, treatment may not be identical to research) - Experimenter effects (biases in delivery of treatment) - Pretest sensitisation (Pre test might tell people what to expect in post test) - Sampling bias (not representative of population) - Random allocation not directly protect external validity, but may help indirectly by protecting internal validity

Question 30

Random selection (sampling) vs random allocation

Answer 30

Random selection: sample is drawn from population by chance Random allocation: Subjects allocated to control or treatment group by chance

Question 31

Experimental designs with random allocation to treatment and control groups give _____ support for causal inferences

Answer 31

best

Question 32

Experimental designs, including best RCTs cant/can prove cause and effect

Answer 32

can't

Question 33

Confidence interval indicates...

Answer 33

the percentage of the time intervals that will contain the population parameter - measures degree of uncertainty or certainty in sampling method