Data Visualization and Cognitive Biases in Research

X-axis

Horizontal axis showing independent variable values.

Y-axis

Vertical axis showing dependent variable values.

Legend

Key explaining symbols/colors in the graph.

Data points

Individual values plotted on the graph.

Pie chart

Displays proportions of categorical data, summing to 1.

Bar plots

Compare values of separate variables, vertical or horizontal.

Error bars

Indicate uncertainty or confidence in averages.

Line graphs

Used for continuous x values, often time-series.

Scatter plots

Show relationships between variables with individual data points.

Histogram

Bar plot representing frequency distribution of data.

Misleading graphs

Graphs that distort data interpretation through manipulation.

Aspect ratio

Width-to-height ratio affecting graph interpretation.

Truncating axes

Cutting axes to exaggerate differences in data.

Magnitude colors

Light to dark shades of the same color.

Divergent colors

Two color shades indicating different directions.

Categorical colors

Different colors representing distinct categories.

Overwhelming data

Too many data points complicating graph readability.

Deception techniques

Methods used to mislead in data presentation.

Manipulation of axes

Altering graph axes to distort data interpretation.

Truncating axes

Cutting off axis scales to exaggerate differences.

Non-linear scales

Using scales that distort proportionality in graphs.

Disproportionate element sizes

Elements sized inaccurately relative to their data.

Selective data presentation

Choosing specific data to support a particular argument.

Overloading with information

Presenting excessive data to confuse the audience.

Visual distractions

Elements that divert attention from key data points.

Guidelines for effective data visualization

Best practices to enhance clarity in data representation.

Simplify graph

Reduce complexity for better understanding of data.

Appropriate encoding

Selecting suitable visual formats for data representation.

Focus on patterns

Highlighting trends or significant details in data.

Meaningful axis ranges

Choosing axis limits that accurately represent data.

Transformations

Mathematical adjustments to improve data visualization.

Visualize density

Represent concentration of data points in scatter plots.

Misleading color choices

Using colors that can misrepresent data relationships.

Consistency across graphs

Maintaining uniformity in style and format across visuals.

Clear labeling

Providing understandable labels for all graph elements.

Audience expertise

Considering the knowledge level of the intended viewers.

System 1 thinking

Fast, automatic, and unconscious thought processes.

System 2 thinking

Slow, deliberate, and conscious reasoning.

Cognitive biases

Systematic errors in thinking affecting judgments.

Heuristics

Mental shortcuts for quick decision-making.

Confirmation bias

Favoring information that confirms existing beliefs.

Representativeness heuristic

Judging probabilities based on similarity to prototypes.

Gambler's fallacy

Misunderstanding independence of random events.

Regression to the mean

Extreme values tend to be followed by average ones.

Availability heuristic

Estimating likelihood based on easily recalled examples.

Anchoring

Relying on initial reference points for decisions.

Decoy effect

Influence of a third option on choice perception.

Loss aversion

Preference for avoiding losses over acquiring gains.

Endowment effect

Valuing owned items more highly than non-owned.

Framing effect

Influence of language on perception of information.

Sunk-cost fallacy

Continuing investment based on prior commitments.

Sample

Subset of a population for statistical analysis.

Sampling error

Variability in estimates due to sample size.

95% confidence interval

Range where 95% of sample estimates fall.

Margin of error

Half of the confidence interval width.

Law of large numbers

Larger samples yield averages closer to true values.

Sampling Error

Variability in sample estimates due to random chance.

Bias

Systematic error affecting survey results, not random.

Probability Sample

Equal chance for all population members to be included.

Random Selection

Choosing respondents randomly from the population.

Weighting

Adjusting survey results to reflect population demographics.

Oversampling

Deliberately including more of certain groups in samples.

WEIRD Problem

Psychological research biased towards Western, educated populations.

Leading Questions

Questions that suggest a particular answer or opinion.

Double Barreled Question

Asking about multiple concepts in one question.

Reactance

Participants' resistance to perceived manipulation in surveys.

Forced Choice Questions

Require respondents to choose one option, enhancing accuracy.

Acquiescence Bias

Tendency to agree with statements, especially among less educated.

Order Effects

Influence of earlier questions on later responses.

Margin of Error

Quantifies uncertainty in poll results due to sampling.

Non-response Bias

Bias from individuals not responding to surveys.

Proofiness

Using statistics to create misleading or false impressions.

Expected Value

Average outcome weighted by probabilities of each outcome.

Curated Coincidences

Creating narratives that highlight specific coincidences.

Poll Weighting

Adjusting results to correct for demographic imbalances.

Predictive Polls

Forecasting outcomes based on survey data.

Voter Prediction Challenges

Difficulties in accurately predicting voter behavior.

Systematic Bias

Consistent error affecting survey results, not random.

Coincidence

An unexpected occurrence prompting new theories.

Cholera Example

London water pump linked to cholera outbreak.

Evaluating Coincidences

Assessing randomness and curator incentives.

Correlation

Relationship between two variables, not causation.

Spurious Correlation

Coincidental relationship without direct impact.

False-Cause Fallacy

X occurring before Y does not imply causation.

Conditional Probability

Probability of one event given another event.

Prosecutor's Fallacy

Confusing P(A|B) with P(B|A), ignoring base rates.

Bayes' Rule

Posterior equals likelihood multiplied by prior.

Likelihood

What is being observed in a probability scenario.

Prior

Existing knowledge or expectations before observation.

Pigeon Hole Principle

More items than containers guarantees shared containers.

Filtering Phenomenon

Focusing on successes while ignoring failures.

Averages and Extremes

Large collections show more extreme values.

Expected Values

Averages weighted by their probabilities.

Rarity

Improbabilities are common; rarity alone isn't evidence.

Oddities in Numbers

Patterns can still contain oddities.

Counter-Intuitive Conclusions

Deviations from mean do not guarantee return to mean.

Regression to the Mean

Trials tend to average out over time.

Poisson Probability Distribution

Predicts rare events based on known rarity.

Flatter Prior Distributions

Indicates less knowledge about prior probabilities.

Flatter Likelihood Function

Indicates less certainty from observations.