14.4 Collecting, Aggregating, and Generating Data: Public Data Sets, Surveys, Sensors, and Crowdsourcing
Key Takeaways
- Publicly available data sets, such as those from data.gov, the U.S. Census Bureau, and NOAA, are free and often large, which lets students and researchers study real questions without collecting data themselves.
- Surveys generate self-reported data: categorical answers, ordered ratings such as 1–5 agreement scales, numeric responses, and open-ended text, all subject to wording, sampling, and nonresponse bias.
- Sensors generate automatic, often continuous numeric measurements over time, such as temperature, motion, light, location, or heart rate, which can be very large and need calibration and cleaning.
- Crowdsourcing collects contributions from large numbers of people online; citizen science applies it to research, as in eBird, Galaxy Zoo, and Foldit.
- Choose a data-collection method that measures what the question actually asks, from a sample that represents the population of interest.
What this competency asks
ETS asks you to be familiar with a variety of computational methods for data collection, aggregation, and generation:
- Identify the benefits of working with publicly available data sets.
- Identify the types of data generated by surveys and sensors.
- Identify examples of crowdsourcing and citizen science.
- Identify appropriate data-collection methods for a given context and purpose.
Publicly available data sets
Governments, universities, and organizations publish data for anyone to use.
| Source | Example data |
|---|---|
| data.gov (U.S. government) | Education, transportation, energy, and health data |
| U.S. Census Bureau | Population, income, and housing by region |
| NOAA and NASA | Weather, climate, ocean, and satellite data |
| World Bank, WHO | International development and health indicators |
| City open-data portals | Transit, 311 service requests, school data |
| Research repositories, Kaggle | Curated data sets for analysis and machine learning |
Benefits
- Free and immediately available: no time or cost spent collecting
- Large and broad: often far bigger than a class or school could gather
- Documented and standardized, with codebooks that describe fields
- Reproducible: others can check an analysis using the same data
- Authentic questions for students, such as local air quality, school enrollment trends, or weather patterns
Cautions: check the license and terms of use (Section 3.1), read the documentation for definitions and units, watch for missing data and changes over time, and consider privacy, since data can sometimes be re-identified when combined with other sources (Section 3.3).
Data from surveys
Surveys ask people to report information about themselves: opinions, behaviors, and characteristics.
| Type of survey data | Example question | Nature of the data |
|---|---|---|
| Categorical (nominal) | "Which language do you speak at home?" | Categories with no order |
| Ordinal (rating scale) | "Rate your agreement from 1 (strongly disagree) to 5 (strongly agree)" | Ordered categories; the gaps between them are not necessarily equal |
| Numeric | "How many hours did you study last week?" | Numbers, but self-reported and possibly inaccurate |
| Open-ended text | "What would improve this course?" | Rich but harder to analyze; needs coding or text analysis |
Limitations to recognize:
- Sampling bias: respondents do not represent the population. An online survey misses people without Internet access.
- Nonresponse bias: the people who choose not to respond differ from those who do.
- Wording and order effects: a leading question ("Don't you agree that…") shapes answers.
- Self-report error: memory lapses and a tendency to give socially desirable answers.
Online forms collect survey data digitally, validate responses (Section 11.2), and export directly to spreadsheets.
Data from sensors
Sensors convert physical conditions into numbers automatically.
| Sensor | Measures |
|---|---|
| Thermometer or thermistor | Temperature |
| Accelerometer and gyroscope | Motion and orientation (step counters, phones) |
| Light, sound, and CO₂ sensors | Environmental conditions |
| GPS receiver | Location over time |
| Optical heart-rate sensor | Pulse |
| Camera | Images and video |
Characteristics of sensor data:
- Objective measurements rather than opinions
- Time series: readings at regular intervals, often continuous
- Large volume: one reading per second is 86,400 readings per day, per sensor
- Needs cleaning: calibration drift, noise, and gaps when a device loses power or connection
Sensors are central to embedded systems and the Internet of Things (Section 15.3).
Crowdsourcing and citizen science
Crowdsourcing gathers work, ideas, or data from many people, usually online. Citizen science applies crowdsourcing to scientific research.
| Project | What volunteers contribute |
|---|---|
| Wikipedia | Encyclopedia articles and edits |
| OpenStreetMap | Map data |
| eBird (Cornell Lab of Ornithology) | Bird sightings with location and time |
| Galaxy Zoo (Zooniverse) | Classifications of galaxy images |
| Foldit | Protein-folding puzzle solutions |
| Community air-quality networks | Readings from home sensors |
Benefits: enormous scale, wide geographic coverage, low cost, and public engagement with science. Challenges: uneven data quality, which projects address with multiple independent labels, expert review, and training; biased coverage (more observations where more volunteers live); and privacy in location data.
Aggregation and generation
- Aggregation combines data from many sources or summarizes many records, such as combining attendance from every school, or averaging hourly readings into daily values. Aggregation reveals patterns but can hide variation within groups, and combining data sets can create privacy risks.
- Data generation also happens without people answering questions: server logs, app usage records, transactions, and simulations (Section 14.2) all generate data.
Choosing an appropriate collection method
| Purpose | Appropriate method |
|---|---|
| Students' opinions of a new schedule | Survey (anonymous; ideally a representative sample) |
| Classroom temperature and CO₂ over a month | Sensors logging at regular intervals |
| Population trends in your county over 20 years | Public data set (Census) |
| Bird migration across a continent | Citizen science (eBird) |
| Whether a new study method improves scores | A designed experiment with randomly assigned groups |
| How long students take to use a new app feature | Automatically collected usage logs, with consent |
Always ask: Does this method measure what the question asks? Does the sample represent the population? Is collecting these data ethical, and do participants consent?
A science class wants to know how carbon-dioxide levels in its classroom change during each school day over a month. Which data-collection method is most appropriate?
A survey item asks students to rate "I feel prepared for the test" from 1 (strongly disagree) to 5 (strongly agree). What kind of data does this produce?
What is a key benefit of having students analyze a publicly available data set, such as census or weather data, rather than collecting their own data?
Which is an example of citizen science?