Data Levels & Descriptive Statistics

Once data has been collected, it has to be described before it can be tested. This lesson equips you with the descriptive-statistics toolkit OCR expects at A-Level — and, importantly, the mathematical skills that make up the compulsory quantitative element of the qualification. Around 10% of the marks across the whole A-Level assess mathematical skills, and most of those live in Component 01. You will be asked to calculate a mean, work out a percentage, read a standard deviation, express a number in standard form, or choose the right graph — sometimes without a calculator prompt.

We cover the levels and types of data (nominal, ordinal, interval; quantitative/qualitative; primary/secondary), the measures of central tendency (mode, median, mean) and dispersion (range, variance, standard deviation — with the formula in full), the numerical skills of ratio, percentages, fractions, standard and decimal form and significant figures, and the family of graphs used to present psychological data. Getting the level of data right is the linchpin: it determines which average, which measure of spread, and (in the next lessons) which statistical test you may use.

Spec Mapping

This lesson covers	OCR H567 Component 01 sub-area	AO focus
Nominal, ordinal, interval data; quantitative/qualitative; primary/secondary	1.3 Data recording, analysis, presentation — levels/types of data	AO1; AO2
Central tendency: mode, median, mean	1.3 — descriptive statistics	AO1; AO2 calculation
Dispersion: range, variance, standard deviation	1.3 — descriptive statistics	AO1; AO2 calculation
Ratio, percentages, fractions; standard & decimal form; significant figures	1.3 — mathematical skills	AO2
Graphs: bar, line, pie, histogram, scatter diagram	1.3 — presentation	AO1; AO2

Referenced descriptively; see the official OCR H567 specification document for exact wording. This lesson develops AO1 (defining data levels and statistics) and, heavily, AO2 (calculating and selecting the correct statistic or graph for given data — the mathematical-skills marks).

---

Levels and Types of Data

The level of measurement describes how much information a set of numbers carries. OCR examines three levels.

• Nominal data are frequency counts in named, unordered categories — e.g. how many participants chose "yes" versus "no", or the number of left- versus right-handers. You cannot order or average the categories; you can only count how many fall in each. Nominal is the least informative level.
• Ordinal data can be ranked in order, but the intervals between ranks are not equal or known — e.g. positions in a race, or scores on a Likert-based attitude scale. A rating of 4 is "more" than 2, but not necessarily "twice as much".
• Interval data sit on a scale with equal, defined intervals between points — e.g. temperature in °C, reaction time in milliseconds, or scores on a standardised test treated as having equal units. Interval is the most informative of the three and is required (along with other criteria) for a parametric test.

A memory aid used by many students: N-O-I runs from least to most information — Nominal (name only), Ordinal (order), Interval (equal intervals).

The reason the level of data matters so much is that it governs what you are allowed to do with the numbers. Nominal data are merely labels that happen to be counted — you can say how many participants fall in each category, but it is meaningless to average them (the "average" of "yes = 1" and "no = 2" has no interpretation), so the only appropriate measure of central tendency is the mode. Ordinal data can be ranked, so a median (the middle rank) is meaningful, but because the intervals between ranks are not equal, arithmetic on the values is suspect — the difference between a Likert "2" and "3" may not equal the difference between "4" and "5", so a mean of Likert scores is technically questionable even though researchers often compute one. Interval data have equal, defined units, so all arithmetic is legitimate, the mean is appropriate, and — provided the other criteria are met — the more powerful parametric tests become available. In short, the level of data is not an academic classification but a permission slip: it tells you which averages, which measures of spread, and which inferential tests are valid. Choosing a statistic that the data level does not support is a genuine error, not a stylistic one.

A frequent source of confusion is that the same underlying phenomenon can be measured at different levels depending on how the researcher operationalises it. "Anxiety" could be recorded nominally (anxious / not anxious), ordinally (rank participants from least to most anxious, or use a Likert self-rating), or on an interval scale (a physiological measure such as heart rate in beats per minute). The level is a property of the measurement, not of the concept — a point that connects directly to the operationalisation decisions of Lesson 2 and 4, and one that examiners test by asking candidates to identify the level of data a described measure produces.

Two further distinctions cut across these:

• Quantitative data are numerical (how much/how many); qualitative data are non-numerical, describing meaning in words or images (e.g. interview transcripts). Quantitative data are easy to analyse statistically but can lose depth; qualitative data are rich but harder to summarise objectively.
• Primary data are collected first-hand by the researcher for the current study; secondary data are pre-existing data collected by someone else for another purpose (e.g. government statistics, another researcher's dataset). Primary data fit the aim exactly but are costly to gather; secondary data are quick and cheap but may not fit the aim precisely.

These cross-cutting distinctions each carry their own evaluative trade-offs, which are examinable in their own right. Quantitative data — numbers — are objective, easy to summarise with descriptive statistics, straightforward to compare across large samples, and amenable to inferential testing; their weakness is a loss of depth, as reducing a rich human experience to a number can strip away meaning and context. Qualitative data — words, images, transcripts — capture that depth and can reveal the why behind a behaviour, but are laborious to analyse, harder to compare, and more exposed to the researcher's subjective interpretation. Many strong studies therefore gather both: quantitative data to establish what happens and how much, qualitative data to illuminate why. Primary data, collected for the specific study, fit the research question precisely and give the researcher full control over how they were gathered, but demand time, money and effort. Secondary data — existing datasets, official statistics, prior research — are fast and inexpensive and can enable studies (e.g. of historical trends) that primary collection never could, but were gathered for other purposes and so may not fit the current aim, may be of unknown quality, and may be outdated. Recognising which type a study used, and weighing the corresponding trade-off, is a routine source of AO3 marks.

graph TD
    A["Level of data"] --> B["Nominal<br/>named categories, counts only"]
    A --> C["Ordinal<br/>can be ranked; unequal intervals"]
    A --> D["Interval<br/>equal, defined intervals"]
    B --> E["Least information → mode; chi-square/sign test"]
    C --> F["Ranked info → median; Wilcoxon/Mann-Whitney/Spearman"]
    D --> G["Most information → mean; parametric tests (if criteria met)"]
    style B fill:#c0392b,color:#fff
    style C fill:#e67e22,color:#fff
    style D fill:#27ae60,color:#fff

---

Measures of Central Tendency

A measure of central tendency summarises a dataset with a single "typical" value.

The mode is the most frequently occurring value. It is the only average usable with nominal data, is unaffected by extreme scores, but may not exist or may be several values.

The median is the middle value when the data are placed in order (the mean of the two middle values if there is an even number). It is appropriate for ordinal data, is unaffected by extreme scores (outliers), but ignores the exact values of most data points.

The mean is the arithmetic average — the sum of all values divided by the number of values:

$$\bar{x} = \frac{\sum x}{n}$$xˉ=n∑x​

The mean uses every value, making it the most sensitive and informative average, and it is required for interval data and parametric tests. Its weakness is that it is distorted by extreme scores — a single outlier can drag the mean far from the "typical" value, which is exactly when the median becomes preferable.

Choosing the right measure of central tendency is a real skill, not a formality, and it depends on both the level of data and the shape of the distribution. For nominal data there is no choice — only the mode applies. For ordinal data the median is the natural summary, since it respects the ranking without assuming equal intervals. For interval data the mean is usually preferred because it uses every value and feeds directly into parametric analysis — but only when the distribution is roughly symmetrical. The moment the data are skewed by extreme scores, the mean is pulled toward the tail and ceases to describe a "typical" case, so the median (which cares only about the middle position, not the values of the extremes) becomes the more honest summary. This is why reporting income, house prices or reaction times — all typically skewed — conventionally uses the median rather than the mean. The mode, meanwhile, has the special virtue of being the only average that identifies the most common value, which can be the most useful summary when the question is "what is typical?" in a categorical sense, though it can be unstable (small changes in the data can shift it) or non-unique (a distribution can have two modes). A strong answer selects the average that suits the data and justifies the choice, rather than defaulting to the mean out of habit.

Average	Uses	Strength	Weakness
Mode	Nominal (and any)	Only average for nominal; ignores outliers	May not exist / not unique
Median	Ordinal+	Unaffected by outliers	Ignores exact values
Mean	Interval	Uses all data; most sensitive	Distorted by extreme scores

---

Measures of Dispersion

A measure of dispersion (spread) describes how varied the data are around the centre. Two datasets can share a mean but differ wildly in spread.

The range is the difference between the highest and lowest value (some conventions add 1). It is quick to calculate but uses only the two extreme values, so a single outlier makes it misleading.

The variance and the standard deviation use every value to measure the average distance of scores from the mean. The standard deviation (SD) is the square root of the variance and is expressed in the same units as the original data, which is why it is the preferred measure of spread for interval data. The formula OCR expects (using $n-1$n−1 for a sample) is:

$$s = \sqrt{\dfrac{\sum (x - \bar{x})^2}{n - 1}}$$s=n−1∑(x−xˉ)2​​

where $x$x is each score, $\bar{x}$xˉ is the mean, and $n$n is the number of scores. In words: find the mean; subtract it from each score and square the result; add those squared deviations; divide by $n-1$n−1; take the square root. A large SD means scores are widely spread (more variability, less consistency); a small SD means scores cluster tightly around the mean (high consistency). The variance is simply the SD squared:

$$s^2 = \dfrac{\sum (x - \bar{x})^2}{n - 1}$$s2=n−1∑(x−xˉ)2​

Reporting a standard deviation alongside a mean is an A*-level habit: it tells the reader not just the typical score but how representative that typical score is of the group.

Why does dispersion matter as much as central tendency? Because two groups can share an identical mean yet be utterly different. Imagine two classes both averaging 50% on a test: in the first, everyone scored between 45 and 55 (small SD — the class is consistent and the mean describes them well); in the second, half scored around 20 and half around 80 (large SD — almost nobody actually scored near the mean, which is deeply unrepresentative). A mean reported without a measure of spread can therefore be positively misleading. The standard deviation quantifies exactly this — the average distance of scores from the mean — and does so in the original units, which is why it is more informative than the range. The range, by contrast, is calculated from only the two extreme values and so is thrown off entirely by a single outlier; it is quick but crude. The variance conveys the same information as the SD but in squared units, which are hard to interpret directly (squared marks, squared seconds), so the SD — its square root — is usually the reported figure. In evaluation, the ability to say "the large standard deviation shows the effect was inconsistent across participants" or "the small standard deviation shows the finding was reliable within the group" is a genuine AO3 move that most candidates neglect.

---

Mathematical Skills for Component 01

The quantitative skills OCR assesses are straightforward but must be secure under exam pressure.

Percentages. A percentage expresses a value as a proportion of 100. To convert a proportion to a percentage, multiply by 100: if 18 of 24 participants helped, that is $\frac{18}{24} \times 100 = 75\%$2418​×100=75%. (Note the escaped percent sign in the working — a bare percent inside maths would break the equation.) To find a percentage of a quantity: 20% of 60 is $0.20 \times 60 = 12$0.20×60=12.

Fractions and ratios. A ratio compares quantities: if a sample has 12 males and 18 females, the ratio is $12 : 18 = 2 : 3$12:18=2:3 when simplified. Fractions, decimals and percentages are interchangeable: $\frac{3}{4} = 0.75 = 75\%$43​=0.75=75%.