Sociological Research

This is Unit Two of Sociology: research. Unit One made a claim that this unit has to cash. Sociology is not common sense, because common sense is untested — which only helps if there is a way to test things. This unit is that way. It opens with a limit rather than a method, and the limit is worth taking seriously. Empirical work can establish what is the case. It cannot, by itself, establish what ought to be. Any study that appears to do both has smuggled something in. The middle is the toolkit — sampling, surveys, fieldwork, experiments — and one error entry inside it corrects something almost everybody believes. A bigger sample does not repair a biased one. How members were selected governs whether a finding generalises, and no amount of size touches that. Then the most practical run in the unit: five procedures for reading numbers, and an error entry saying a chart never establishes a claim — it only displays one, through design decisions somebody made. By the end you will be able to do five things. One: say what value neutrality claims and what it does not. Two: tell reliability from validity. Three: say why a bigger sample cannot fix a biased one. Four: read a rate, a cross-tab, a trend line and a distribution. Five: say what a chart establishes.

Three entries. What research is for, the sequence it follows, and then a boundary — because a study can tell you what is happening and cannot, on its own, tell you what to do about it.

What it replaces.

Soc 0141 Why Sociological Research

Replacing impression with evidence, using procedures designed so another researcher could repeat the work. Repeatable by someone else is the whole standard.

The sequence.

Soc 0142 The Scientific Method in Sociology

Ask a question, review prior work, form a hypothesis, collect data, analyse, report. Fixed order — each step depends on the one before it.

And the limit.

Soc 0143 Value Neutrality

Empirical science can establish what is the case and cannot by itself establish what ought to be. A study that seems to deliver both has brought a value in from outside and not declared it.

Four entries on the work before any data exists. Narrowing a topic until it is answerable, finding out what is already known, and committing to a prediction — in that order, because the order is what stops a study from being written backwards from its result.

The question.

Soc 0145 The Research Question

Narrowing a topic into something a study can answer, specifying who is studied and what is measured. Most topics are not questions yet.

What is already known.

Soc 0146 The Literature Review

Surveying published research to establish what is known, what is disputed, and what is missing. The third category is where a new study earns its place.

The prediction.

Soc 0148 The Hypothesis

A testable statement predicting a relationship between factors, written so evidence could contradict it. Could contradict is the requirement.

The two factors.

Soc 0149 Independent and Dependent Variables

The independent variable is treated as the cause; the dependent variable is expected to change in response. Naming which is which names the direction of the claim.

Three entries, and the last two are a pair that get confused constantly. A measure can be perfectly consistent and measure the wrong thing — consistent and wrong at once, which is worse than noisy.

Making it countable.

Soc 0151 Operational Definitions

Stating exactly how an abstract concept will be measured in this study, turning something arguable into something countable.

The first property.

Soc 0152 Reliability

Producing the same result on repeated application to the same thing under the same conditions. Consistency, and nothing more than consistency.

The second.

Soc 0153 Validity

Capturing the concept it claims to capture rather than something merely correlated with it. A reliable measure of the wrong thing is still the wrong thing, delivered dependably.

Two entries. Numbers and words — and the useful framing is that these answer different questions rather than being better or worse versions of each other.

Numbers.

Soc 0155 Quantitative Data

Represented as numbers, letting a researcher count, compare and detect patterns across large groups. Scale is what it buys.

Words.

Soc 0156 Qualitative Data

Represented in words, images and observations, capturing meaning, context and process rather than magnitude. It answers how and why where the other answers how many.

Three entries, ending in the error entry most worth carrying out of this unit. A poll of a hundred thousand people selected badly tells you less than a poll of a thousand selected well, and the intuition runs hard the other way.

The gap.

Soc 0158 Population and Sample

The population is everyone a study aims to describe; the sample is the smaller group actually studied and generalised from. The claim is about the first.

Building it well.

Soc 0159 The Random Sample

Probability sampling gives every member a known, nonzero chance of selection. That known chance is what licenses generalising at all.

And the error.

Soc 0160 "A Bigger Sample Is a Better Sample"

Sample size cannot repair a biased sample. How members were selected governs whether a finding generalises — adding more of the same wrong people makes the error more confident, not smaller.

Two entries on collecting answers directly. Standardised questions so responses can be compared — and then the trade between letting somebody answer alone and having a researcher present to follow up.

At scale.

Soc 0162 The Survey

Collecting standardised answers from many respondents so responses can be compared and counted. Standardisation is what makes comparison possible.

Two ways.

Soc 0163 Interviews and Questionnaires

A questionnaire is completed alone; an interview has a researcher who can follow up. The follow-up buys depth and costs comparability, which is the trade in one line.

Four entries on methods that go to the setting rather than summoning people out of it, ordered by how far in the researcher goes — from watching, to joining, to living there long enough to describe a whole way of life.

The category.

Soc 0165 Field Research

Studying people in their natural setting rather than a laboratory or a questionnaire. What is gained is behaviour in its own context.

Going further.

Soc 0166 Participant Observation

Joining the group being studied, taking part while systematically recording. Participation changes what can be seen, in both directions.

Furthest.

Soc 0167 Ethnography

Extended study of a whole way of life, immersing long enough to describe a group on its own terms. Time is the method.

And one deep case.

Soc 0169 The Case Study

A single person, group, event or organisation examined in depth, trading breadth for detail no wider study could reach. The trade is explicit rather than a weakness.

Four entries. The design that licenses a causal claim, the comparison that makes it work — and then two methods that collect no new data at all, which is a genuine option rather than a fallback.

The design.

Soc 0170 The Experiment

Randomly assigning subjects to conditions and changing one factor while holding others constant. The random assignment is what makes the comparison mean anything.

The comparison.

Soc 0171 Control and Experimental Groups

The experimental group receives the treatment; the control group does not, and exists to show what happens without it. Without the second there is nothing to compare to.

Data already gathered.

Soc 0173 Secondary Data Analysis

Using data somebody else already collected — a census, an agency, an earlier study — to answer a new question. Scale and history, for free.

And recorded communication.

Soc 0174 Content Analysis

Studying recorded communication systematically, coding texts, images or broadcasts to count patterns. It turns culture into something countable.

Two entries. Two things moving together, and the inference that does not follow — which is the single most common misreading of research anywhere, not only in sociology.

The pattern.

Soc 0176 Correlation

A measured pattern in which two factors vary together, in the same direction or opposite ones. A description of movement, and nothing more.

And the error.

Soc 0177 "Correlation Proves Causation"

Two factors varying together does not establish that either causes the other, because a third may drive both. The pattern is real; the explanation is not in it.

Six entries, and this is the most portable run in the course. Five reading procedures — rate, cross-tabulation, trend line, distribution, and a chart built to mislead — and then an entry stating what a chart can and cannot do.

Watch a count become a rate.

Soc 0178 Reading a Rate, Not a Count

Converting a raw count into a proportion of the group it came from, because counts follow population size before they follow anything interesting. The bigger place has more of everything.

Watch a table get read.

Soc 0179 Reading a Cross-Tabulation

Finding the relationship between two variables by comparing percentages within the table rather than reading the raw cells.

Watch a trend get checked.

Soc 0180 Reading a Trend Line

Establishing what is measured, over what period, against what baseline, before accepting the direction the line appears to show. All three, before the shape means anything.

Watch a mean get tested.

Soc 0181 Reading a Distribution and Its Skew

Asking how values spread around the average, because a mean describes a symmetric spread well and a skewed one badly. A single number hides the shape it came from.

Watch a chart get audited.

Soc 0182 Reading a Misleading Chart

Checking the axis, the baseline, the interval and the scale before reading the picture. Four decisions, each of which can change the impression without changing the data.

And the error.

Soc 0183 "The Chart Proves It"

A chart displays data and never establishes a claim, because every design decision behind it was made by somebody. Persuasiveness is a property of the drawing.

Three entries closing the unit. A professional code, the participant's agreement, and a committee whose approval is required before any data is collected — not reviewed afterwards.

The code.

Soc 0184 The ASA Code of Ethics

Binding researchers to professional standards covering competence, integrity and the treatment of participants. Enforceable within the profession.

The agreement.

Soc 0185 Informed Consent

Participants are told what the study involves and any risk it carries, and agree to take part on that basis. Told first — agreement without information is not consent.

And the gate.

Soc 0186 The Institutional Review Board

A committee reviewing proposed research involving human participants before it begins. Before is the operative word — this is not an inquiry into what happened.

Five things you can now do. You can say what value neutrality claims — that empirical work establishes what is and not what ought to be — and that a study appearing to do both has brought a value in without declaring it. You can tell reliability from validity, and say why a measure can be perfectly consistent while measuring the wrong thing. You can say why a bigger sample cannot repair a biased one: selection governs generalisation, so more of the same wrong people makes the error more confident rather than smaller. You can read a rate rather than a count, a cross-tabulation by percentage, a trend line against its baseline, and a distribution behind its mean. And you can say what a chart does — it displays data and never establishes a claim, because every decision behind it was somebody's.