Week 1 Lecture 5: Fraud & Replication; Frequency Data

Flashcards cover fraud vs. error, vaccine case study, scientific rigor, power-pose replication, replication types, frequency data and validity, measurement scales, survey question formats, Likert scales, major question-writing mistakes, and ethical scientific writing.

What are the two main ways false information can enter the scientific literature, even with ethics boards in place?

Straight-up fraud and unintentional errors.

Why might straight-up fraud occur in science?

Because of personal beliefs, publication incentives, or financial conflicts of interest.

What are typical causes of unintentional errors in research?

Sloppy or uninformed practices and simple bad luck.

What safeguards help catch fraud or errors in science?

Peer review and replication.

Give one consequence of scientific fraud for public trust.

It creates public distrust that can reduce acceptance of valid science (e.g., vaccine hesitancy).

How contagious is measles?

Highly contagious and can lead to serious complications.

Name two serious complications of measles mentioned in lecture.

Pneumonia and encephalitis.

What percentage effectiveness do two doses of MMR vaccine provide against measles?

About 97% effective.

Which 1998 journal article falsely linked MMR vaccine to autism?

The retracted Wakefield et al. paper in The Lancet.

Despite retraction, why is the MMR–autism myth still influential?

It was highly publicized and remains strong in collective imagination.

What three groups share responsibility for ensuring scientific rigor?

Researchers, the scientific field, and science consumers.

List one open science practice researchers can use to promote transparency.

Pre-registering studies or sharing data and materials openly.

What was the central claim of Amy Cuddy’s original power-pose study?

Two-minute high-power poses change hormones, risk taking, and feelings of power.

Name one physiological variable power-pose researchers claimed was altered.

Testosterone (increase) or cortisol (decrease).

What did later large-scale replications conclude about power posing?

It increases feelings of power but shows no reliable behavioral or hormonal effects.

Define direct (exact) replication.

Repeating a study with the same methods to see if the results recur.

Define conceptual replication.

Testing the same hypothesis with different methods or measures to assess generalizability.

Why are replications important?

They inform us about reliability and generalizability of scientific effects.

What type of claim is based on measuring the frequency of a single variable?

A frequency claim.

Which two validities are most relevant when evaluating a frequency claim?

Construct validity and external validity.

In frequency data, is the variable manipulated or measured?

Measured.

What is the difference between categorical and quantitative variables?

Categorical variables have category levels; quantitative variables are coded with numbers representing amounts or ranks.

Give an example of an ordinal variable.

Finishing order in a race.

Give an example of an interval variable.

IQ score (equal intervals, no true zero).

Give an example of a ratio variable.

Number of exam questions answered correctly (has a meaningful zero).

What are open-ended survey questions?

Questions allowing respondents to answer in their own words.

State one pro and one con of open-ended questions.

Pro: rich, spontaneous answers. Con: difficult to code.

What are forced-choice questions?

Questions that ask respondents to select from listed answer options.

State one pro and one con of forced-choice questions.

Pro: easy to code/analyze. Con: answer choices may not capture true opinions.

What is a Likert scale?

A rating scale where respondents express degree of agreement with statements, typically 1 = Strongly Disagree to 5 = Strongly Agree.

Why might including a neutral midpoint on a Likert scale matter?

It allows ambivalent respondents a middle choice but can encourage fence-sitting.

List the five common mistakes to avoid when writing survey questions.

1) Asking about things participants can’t know or recall accurately, 2) Leading or loaded wording, 3) Double-barreled questions, 4) Double negatives, 5) Confusing or ambiguous wording.

Provide an example of a leading question.

“Do you support the pro-life view that abortion is murder?”

Provide an example of a double-barreled question.

“I am generally a very relaxed and reliable person.”

Why are double negatives problematic in survey items?

They are confusing and prone to misinterpretation (e.g., “I am not unexcitable”).

What type of measurement scale is popularity rank when students are ordered from most to least popular?

Ordinal scale.

Which big validity is least relevant for simple frequency claims?

Internal validity (they do not assert causality).

What is the ‘Mandela Effect’ study an example of?

Using frequency data to show shared false memories across people.

What do Google Trends frequency graphs (e.g., ‘adopt a dog’ searches) illustrate?

Changes in search behavior over time for a single variable.

What does the ‘dived vs. dove’ Ngram example demonstrate about frequency data?

It can reveal linguistic trends and regularization of verb forms across decades and dialects.

Explain margin of error in the context of frequency claims.

A confidence interval indicating precision of an estimated percentage from a sample.

What does ‘construct validity’ ask in survey research?

Whether the question truly measures the intended concept.

What does ‘external validity’ ask for frequency data?

To which populations, settings, and times the estimate can generalize; depends on sample representativeness.

Why must scientific writing include citations?

To give credit, avoid plagiarism, and allow readers to verify sources.

How can ChatGPT be used ethically in academic work?

As a tool for idea generation or editing while always citing sources and not passing AI-generated content as entirely original work.

When rewriting a poor survey question, what should be specified for a forced-choice version?

Clearly defined response options and scale labels.