research design and experimentation II

Question 1

all data…

Answer 1

• misbehaves
• not all can be randomised

Question 2

RCTs may have

Answer 2

• good internal validity (minimal confoundment, captures relevant variables)
  -lack external validity (not transferable across populations)
• restricts generalisability

Question 3

control variables

Answer 3

• can account for confoundment
• if B1 doesn’t change, not properly randomised
• mitigate effects of unobservables in quasi-experiments

Question 4

quasi-experiments

Answer 4

• experiment not ethical or possible
• facilitated by data availability
• difficult to control factors
• randomisation is accidental, and is not allowed by type of data

Question 5

difference in difference

Answer 5

• T = time (0 before, 1 after)
• G = binary variable of treatment effect (0, 1)
• B3 = T X G = interaction term = DID = treatment effect

Question 6

treatment effect formula

Answer 6

B0 + B1time + B2treatment + B3time x treatment + B4control + error

Question 7

time effect

Answer 7

natural change in outcome overtime

Question 8

group effect

Answer 8

natural change in outcome because of group not attributed to intervention/treatment

Question 9

treatment effect

Answer 9

change in outcome attributed to intervention/treatment

Question 10

interaction term

Answer 10

DID; nets out time and group effect