Lecture 18 ARM

Question 1

Boxplot as a 5 number summary

Answer 1

1. Minimum value (whisker)
2. First quartile (Q1)
3. Median (middle line)
4. Third quartile (Q3)
   5.Maximum value (whisker)

Min and max here are not including the outliers - they are plotted outside of the boxplot itself as a cheeky dot

Question 2

Probability in research

Answer 2

1. Science and uncertainty: all scientific claims are probabilistic
2. Probability basics: predicting likelihood of events
3. Anthropology context: probability helps quantify uncertainty in our findings
4. From probability to sampling: Enables us to generalise from sample to population
5. Goal: Use probability to make informed inferences, NOT absolute proofs - they never say it is a 100% accurate

Question 3

Why do we sample?

Answer 3

1. It is impossible to study everyone: Too large/dispersed populations
2. Resource limits: Time, money, personal constraints
3. Access and ethics: Some are hard to reach, or sensitive to involve
4. Less intrusive: Sampling can reduce burden on the community that is studied - not everyone is studied or involved
5. Speed and feasibility: Enables timely research results by focusing on fewer

Question 4

Representative sample

Answer 4

Reflects population’s diversity and key characteristics

Eg half men and women, spread of ages, jobs etc

Question 5

Sampling bias

Answer 5

Systematic over- or under-representation of some group or trait - occurs usually based on how you choose to sample

Eg surveying only market-goers.
Consequences: Biased sample = skewed findings - not true for population
Anthropology impact: misrepresentation of a community can lead to flawed or even harmful conclusion eg stereotypes

Question 6

Probability sampling (random selection)

Answer 6

Every member has a known, non-zero chance of selection

Question 7

Simple-random sampling

Answer 7

Pure-lottery method - eg random number generator picks people

Question 8

Stratified sample

Answer 8

Divide population into subgroups, sample then randomly within each groups - eunsure representation of key categories eg gender, age, ethnic groups

Question 9

Cluster sampling

Answer 9

Sample groups or clusters (eg villages, households), then individuals within clusters

Usually national surveys or urban - can create a bit dispersed results if areas vary a lot

Question 10

Systematic sampling

Answer 10

Select every n-th individual from a list after a random start.
Eg every second person, every 10th

Question 11

Non-probability sampling (when random is not possible) (!)

Answer 11

If random selection is not feasible, possible for whatever reason

1. Convenience sampling
2. Snowball sampling
3. Purposive sampling

Pros: Easier and sometimes the only option
Cons:Higher risk of bias and limited generalisability.

Anthro reality: Common in ethnography due to practical constraints and requires careful interpretation

Question 12

Convenience sampling

Answer 12

Selecting whoever is readily available (ease over randomness)

Question 13

Snowball sampling

Answer 13

Participant recruit others - useful for hard to reach groups

Possible to capture social networks, bounce of others words in new interviews

Question 14

Purposive sampling

Answer 14

Deliberately selecting individuals for their knowledge or characteristics

Question 15

Example - Program Evaluation (from reading, Bennett &Hays 2023)

Answer 15

Study focus: Civic engagement program EYPC for youth in Illinois

Population = all youth in the program
Sample = participants who took surveys voluntarily

Sampling method: Non-probability (program-based)

Implication: Results apply to engaged youth in program - not necessarily all youth in general

Key finding: Participants showed increased teamwork, leadership etc after program

Question 16

Sample sizes importance

Answer 16

1. General rule is that a larger sample size produces more reliable estimates (aka less random error)
2. Stability - small sample sizes can give wild results ! large samples smooth out the outliers
3. Diminishing returns: beyond some point, doubling sample yields smaller gain in precision. going from 10-100 makes a huge difference. but going from 100 to 200 makes less! at some point, increasing the N might not help that much

Anthro context : often small-N studies - recognise that results may be tentative, cannot make broad generalisations

Key idea: Larger N reduces the influence of random luck who was sampled

Question 17

Margin of error

Answer 17

Definition: The radius of confidence interval; an estimate of how far off the sample result might be from the true population value

Interpretation: “ + or - x% or y units” - range around the sample statistic likely to include the population parameter

Driven by sample size:Larger sample = smaller margin of error aka more precision

Example: Survey result 60% + or - 5% means true population value could be 55% to 65%

Given the sample size, the true population percentage is likely within a percentage range. Eg “between 55-65% support this policy”

Use in anthro: Rare in ethnography, but common in surveys- conveys result reliability (important for presenting quantitative findings responsibly)

Key: this is where chance come into play when we sample - why it is important so see the sampling methods and how it is applied to larger population

Question 18

Confidence interval very easily understood

Answer 18

The population value minus margin of error

Eg population value: 80%
Margin of error: +/- 10 %

Confidence interval = 70% - 90%

Question 19

Confidence Intervals elaborated

Answer 19

Confidence interval (CI): A range of values, derived from the sample, that is likely to contain the TRUEpopulation value

Confidence level: typically 95% in social sciences. this means that if we sampled 100 times, about 95% of those CI would contain the true value

Interpretation: “We are 95% sure that the true mean/proportion lies between x and y.”

Significance: If a CIrange is narrow = precise knowledge.If CIrange is wide = estimate is uncertain.