WEEK 4 L7.2 data collection: case selection

Question 1

two crucial elements of data selection

Answer 1

1. identifying full sets of units (universe of cases)
2. selecting a subset/ sample of data units from that universe

Question 2

key difference between case selection and sampling,

Answer 2

case selection is done deliberately, and involves a small number of cases (controlled by researcher)
probabalistic (e.g random) / non probabalistic

Question 3

two good approaches to small n case selection

Answer 3

1. most similar systems design: similar on most characteristics, but different on the outcome. - what causes the outcome?
2. most different systems design: different on most characteristics, but the outcome is similar

=> selection starts with a causal factor.

Question 4

a bad approach to small n case selection

Answer 4

selection by outcome: general characteristics are different, and cause and effect should both be smilar
- is based on succesful cases, that leads to selection bias
- evidence for the causal mechanism isn’t conclusive because other cases aren’t considered

Question 5

summary of three approaches to small n case selection

Answer 5

1. most similar systems design:
   - gc: similar
   - iv, dv: different
2. most different systems design
   - gc: different
   - iv, dv: similar
3. selection by outcome
   - gc: different
   - iv, dv: similar

Question 6

challenges of case selection 6

Answer 6

1. selection bias
**2. outliers **(if not recognized, might led to an incorrect picture)
3. heterogeneity of cases: comparing cases that are very similar, for example qatar and russia
4. joint history/ historical contingency: a geographical area might be experiencing events, for example world wars- may lead to incorrent inferences
**5. path dependency- **once a system is chosen, it can’t be easily reversed, for example, if one studies health care systems they should recognize this path dependency

Question 7

other tecniques of case selection

5

Answer 7

1. typical case: representative of the universe of cases (on a simple regression line)
2. diverse cases: capturing full variation of cases is captures (putting a grid, and selecting from each square)
3. extreme cases: choosing cases from the outer margins
4. deviant: different from other cases, but not necessarily extreme (move away from the regression line)
5. influential: difficult to show with a plot, not typical, but represent a unique combination of factors that have a large influence on the observed relationship.
6. most similar/ different cases

Question 8

when is ‘typical’ case selection useful?

Answer 8

useful for theory testing
representativeness: yes

Question 9

when is ‘diverse cases’ useful?

Answer 9

useful for theory generating and theory testing
representativeness: maybe
- for example, busy squares can be under represented

Question 10

when is ‘ extreme cases’ useful?

Answer 10

theory generatig (adding an explanation to why they’re unusal)
not representative

Question 11

when is ‘deviant cases’ useful?

Answer 11

theory generating (like extreme cases)
usually not representative

Question 12

what’s an influential case selection?

Answer 12

difficult to show with a plot, not typical, but represent a unique combination of factors that have a large influence on the observed relationship.

Question 13

when is ‘influential cases’ used?

Answer 13

theory testing (influential cases should confirm a theory) but since they’re not typical,they’re not representatives.

Question 14

for what purposes can mdsd and mssd be used for?

Answer 14

theory generating and theory testing
may be representative
focus is more on testing (external validity isn’t the priority)

Question 15

two other tecniques for case selection

Answer 15

1. crucial
2. pathway: cases that allow the researcher to isolate a particilar causal mechanism from alternative mechanisms, and to clearly demonstrate that this cause and nothing else is efficient to cause the effect.
   - cases that are exceptionally relevant to confirm, or disconfirm a theory.
   - so something for theory testing