1.3 - Levels and types of data
Learn what a recorded score represents, how numerical and descriptive information differ, and why the origin of data matters when choosing an analysis.
Categories and ordered scores
A number is only useful if we know what it represents.
Nominal data place observations in named categories without a numerical order. A researcher might record a participant's preferred study setting as quiet, music or either. Coding these as 1, 2 and 3 helps enter them into a spreadsheet; it does not make music twice as much as quiet. Count the frequency in each category. Categories should fit the research question and be clearly defined; ambiguous or overlapping categories can make coding inconsistent.
Ordinal data have a meaningful order, but equal differences between scores are not established. A satisfaction scale from 1 (very dissatisfied) to 5 (very satisfied) orders responses; the psychological gap between 1 and 2 need not equal the gap between 4 and 5. Race positions also order competitors without showing the size of time differences. Ranking scores converts their order into ordinal information and discards precise distances between them.
For example, response times of 200, 210 and 400 milliseconds have ranks 1, 2 and 3 when shortest gets rank 1. Ranks 1 and 2 differ by one rank, just as ranks 2 and 3 do, despite time differences of 10 and 190 milliseconds. Equal raw values share the mean of the rank positions they occupy: two equal lowest scores occupying positions 1 and 2 both receive rank 1.5, and the next score receives rank 3.
Equal measurement intervals
Interval data use a scale on which equal numerical differences represent equal measured differences.
A response time recorded accurately in milliseconds has equal units: the difference between 200 and 210 ms equals the difference between 400 and 410 ms. This supports meaningful subtraction and averaging. H567 uses nominal, ordinal and interval as its three required levels; measurements with equal units and a meaningful zero, such as time or mass, also satisfy its interval-level requirement. In fuller statistical terminology these are ratio measurements. Do not confuse this extra classification with a ratio calculated from category frequencies.
Compare a timed task with a rating of how slow it felt. Both may contain numbers, but a stopwatch supplies equal time units, while a person's 1–5 rating does not establish equal psychological intervals. Similarly, a numerical questionnaire total does not become interval merely because many ratings were added: treating its intervals as equal requires justification from the measurement scale.
Higher measurement detail permits more analysis but does not guarantee validity. A precisely timed task might measure motor speed as well as the intended attention process. Select a measurement that represents the construct, then choose an analysis suitable for that measurement.
Numbers, meanings and coding
Quantitative data express quantities, counts or numerical scores, whereas qualitative data record descriptive meanings and experiences.
“I lose concentration because the lyrics pull my attention away” is qualitative interview material. The number of words recalled in a task is quantitative. Nominal category frequencies are quantitative counts even when their category names are words. Qualitative is therefore not a synonym for nominal, and quantitative is not a synonym for interval.
Numerical summaries make group comparisons efficient and allow statistical analysis. However, a score can omit the reasons behind a response. Detailed interview accounts can reveal unexpected experiences and preserve context, but comparing them requires interpretative choices and may take longer. Neither type is automatically objective or valid: biased questions can distort numbers, and careful qualitative coding can be systematic.
A researcher can turn qualitative accounts into quantitative information through content analysis. Define a coding category such as “mentions distraction by lyrics”, state what counts, and count participants whose accounts contain it. Keep the original accounts so ambiguous decisions can be examined, and use independent coders to check consistency. The resulting frequency answers “how often was this theme recorded?”; it no longer conveys every detail of the original experience. A frequency based on a poorly defined category may look precise while measuring inconsistently.
Where the data came from
The primary–secondary distinction concerns who collected the data and for what investigation.
Primary data are collected first-hand for the current research purpose, for example a psychologist's new interviews about study habits. Secondary data already exist and are reused, for example another research team's interview archive or a published survey dataset. Both can be qualitative or quantitative: an archived transcript is secondary qualitative data; new stopwatch measurements are primary quantitative data.
Primary collection lets the researcher tailor the questions, definitions and procedure to the aim, but it takes time and resources and introduces responsibilities to participants. Secondary data can make a large or historical comparison feasible without repeating collection. However, the original sample, measures or missing information may not fit the new question. Check how the records were produced, their completeness, and whether access and reuse protect confidentiality and follow the original consent conditions.
A researcher examining changes in revision habits might compare new interviews with an older archive. A difference in coded themes could reflect historical change, but also different questions or samples. Data origin must therefore inform the conclusion, rather than treating every available record as directly comparable.