Lecture notes: Likert scales, ethics, data concepts, and generalizability

Research often relies on numerical data to quantify findings. A common method is the Likert scale.
Hospital example: nurses ask patients to rate pain on a chart from 0 to 10 (Likert-type scale).
Survey example: customer service satisfaction from 0 to 5 (or other ranges). Personal anecdote: a call center agent emphasized wanting all fives to boost his review/raise.
Purpose: convert subjective experience (pain, satisfaction) into numerical data for analysis.
Qualitative vs quantitative contrast (foreshadowed): quantitative uses numbers and scales; qualitative is narrative data (e.g., interviews). Both are valid approaches.

Qualitative research: narrative data, often collected via interviews or open-ended responses; data may be stored in journals or transcripts rather than spreadsheets.
Quantitative research: numerical data that can be analyzed statistically (e.g., Likert scores, counts, measurements).
Example of qualitative research: interviewing survivors of a disaster (e.g., 9/11) to understand experiences.
Both types have valid uses; the key is understanding the difference and choosing the appropriate method for the research question.

APA = American Psychological Association; in the US, researchers follow ethical guidelines when involving humans or animals.
Primary focus on human research for test questions, though animal ethics are also important.
Core ethical principles mentioned:
- Obtain informed consent: researchers must inform participants that they are participating in research.
- Assent: parental/guardian permission for a child’s participation (assent is the child’s agreement; not just parental consent).
- Minimize harm: protect participants from greater than usual harm and discomfort; research should avoid causing trauma.
- Confidentiality: protect participants’ data; use pseudonyms or anonymize identifiers (e.g., Jeanie/Jeannie example).
- Debriefing: after participation (especially if deception was involved), explain the study’s purpose and any deception used.

Informed consent: you cannot study people without their knowledge and agreement.
Assent: parental approval for minors to participate; not the same as consent, but a crucial part of ethical participation.
Deception and debriefing: if deception is used, participants must be debriefed afterward to explain the true purpose and the deception.
Confidentiality example: in famous studies, participants’ identities are kept confidential; real names are replaced with placeholders to protect privacy.
Real-world anecdote used to illustrate debriefing: a mock IQ test scenario where the researcher later reveals the true purpose (to study guilt and hand-washing behavior after cheating).

Correlation: a relationship between two variables; can be positive, negative, or no correlation.
Statistical significance (as discussed by the speaker): a finding is significant when it is very unlikely to have occurred by chance.
- The speaker describes it informally as: "less than a 0.05% chance that this occurred by random occurrence." (Note: in many settings, the conventional threshold is 0.05, i.e., 5%. The speaker’s wording reflects a common educational simplification.)
Why significance matters: helps determine whether observed relationships are likely to generalize beyond the sample.

Central tendency measures:
- Mean (average): $\bar{x} = rac{1}{n} \sum{i=1}^{n} xi$
- Median: middle value when data are ordered.
- Mode: most frequent value.
Skewed data: a distribution that is not symmetrical; data lean toward one side.
- Example: hotel guest incomes with a few very high earners can pull the average up, giving a misleading sense of typical income.
Range: difference between the smallest and largest values.
- $\text{range} = \max(x) - \min(x)$
Standard deviation: a measure of how spread out the values are around the mean; indicates variability.
- Population SD: $\sigma = \sqrt{\frac{1}{n} \sum{i=1}^{n} (xi - \mu)^2}$
- Sample SD: $s = \sqrt{\frac{1}{n-1} \sum{i=1}^{n} (xi - \bar{x})^2}$
Normal distribution / bell curve:
- In many domains, data approximate a normal curve where
- Mean (centre) = the average value (e.g., IQ mean = 100 in the example).
- About 68% of data fall within ±1 standard deviation of the mean: $P(\mu - \sigma \le X \le \mu + \sigma) \approx 0.68$
- 100 is the average IQ in the example; the curve is often called the normal curve (also colloquially the bell curve).
IQ example specifics (as mentioned):
- Mean IQ = 100.
- A large portion of the population sits around this mean; a small number have very high or very low IQ scores.
- A joke example: claiming an IQ of 160 would place you in a tiny, rare tail of the distribution; someone with IQ 72 is notably below average and has legal/ethical implications in some contexts.

Generalizable findings: applicable to the broader population beyond the study sample.
Non-generalizable findings: limited to the specific group studied (e.g., a unique island population whose traits may not reflect humanity at large).
Representativeness of the sample: the participants should reflect the diversity of the population of interest (i.e., a representative sample).
More cases generally improve generalizability: larger and more varied samples tend to yield more generalizable conclusions.
Example illustrations from everyday observations:
- Amazon product reviews: a five-star rating with only 19 reviewers is suspicious and may not generalize well to broader user experience.
- A contrasting example: a product reviewed by real users in 1980 with more reviewers despite fewer stars may feel more trustworthy due to broader participation.

The connection between sleep and memory: a correlation example discussed to illustrate how findings can be generalized if the sample is representative and the effect size is robust.
Ethical considerations permeate methodological choices, data handling, and interpretation of results.
The importance of avoiding sensational or misleading statistics in everyday claims and consumer information (e.g., product reviews and self-reported data).

Quick glossary of terms mentioned:

Likert scale: a numeric scale used to represent attitudes or perceptions (often 0–5, 0–10, etc.).
Informed consent: participants’ voluntary agreement to participate after being informed about the study.
Assent: minor participants’ affirmative agreement, typically given by a parent/guardian.
Debriefing: post-study explanation of the study’s purpose and any deception used.
Confidentiality: protection of participants’ identities and data.
Generalizability: applicability of study findings to the broader population.
Representative sample: a sample that reflects the characteristics of the population.
Correlation: a relationship between two variables that can be positive, negative, or none.
Statistical significance: the likelihood that observed results are not due to chance, often discussed in terms of a probability threshold.
Normal (bell) curve: a symmetric, unimodal distribution often used as a model for many natural phenomena.
Central tendency: a center-most value of a data set (mean, median, mode).
Variability: how spread out the data are (e.g., range, standard deviation).
Assent, consent, debriefing, and deception: key ethical components of conducting human research.