Introduction to Data Collection
Data collection is the systematic process of gathering information from the world around us in order to answer questions, test hypotheses, or discover patterns. Every statistical study — whether it investigates the effectiveness of a new medicine, the average screen time of teenagers, or the quality of products on an assembly line — begins with this fundamental step. The way data is collected determines what conclusions can fairly be drawn, and even small errors or biases at this stage can invalidate an entire study. There are two broad categories of data sources: primary data, which you collect yourself directly for a specific purpose, and secondary data, which has been collected by someone else and already exists in databases, reports, or published research. Methods of collecting primary data include surveys, experiments, observational studies, and interviews. Each method has distinct strengths and weaknesses, and choosing the wrong method for a given research question is one of the most common errors in statistical practice. Understanding data collection also means understanding validity — whether a study actually measures what it claims to measure — and reliability — whether repeating the study would yield consistent results. Both concepts depend heavily on how data is gathered, from whom it is gathered, and under what conditions. A beautifully analysed dataset built on flawed collection will produce misleading conclusions, a principle sometimes summarised as 'garbage in, garbage out.'
- Defining the Population and Parameter of Interest
- Primary vs Secondary Data
- Surveys and Questionnaires
- Observational Studies vs Experiments
- Sampling Methods
- Bias in Data Collection
Defining the Population and Parameter of Interest
Before any data is collected, you must clearly define the population — the entire group you want to draw conclusions about — and the parameter you wish to estimate, such as a mean (μ) or proportion (p). For example, if you want to know the average sleep duration of Year 11 students in your country, the population is all Year 11 students nationally, and the parameter is the population mean μ. Collecting data from only one school without accounting for this mismatch leads to conclusions that may not generalise.
Primary vs Secondary Data
Primary data is collected first-hand by the researcher: you design a survey, run an experiment, or observe subjects directly. It is tailored to your exact question but is time-consuming and costly. Secondary data comes from existing sources such as government census records, academic journals, or company databases. It is quick and cheap to access but may not match your question precisely and you have no control over how it was originally collected. A critical question for secondary data is always: 'Was this data collected in a way that is appropriate for my current research question?'
Surveys and Questionnaires
Surveys involve asking a sample of people questions, either in writing (questionnaire) or verbally (interview). Key design considerations include: avoiding leading questions (questions that hint at a desired answer), using clear and unambiguous wording, and deciding between open-ended questions (richer data, harder to analyse) and closed questions (easier to quantify). For example, 'Don't you agree that social media is harmful?' is a leading question. A neutral version would be: 'What is your view on the impact of social media on mental health?'
Observational Studies vs Experiments
In an observational study, the researcher records what naturally occurs without intervening. This can reveal correlations but cannot establish causation because there may be confounding variables — hidden factors that influence both the explanatory and response variables. In a controlled experiment, the researcher deliberately manipulates an independent variable and randomly assigns subjects to a treatment group or a control group, allowing causation to be established. The use of random assignment is critical: it distributes the effects of confounding variables approximately equally across groups, making any observed difference attributable to the treatment.
Sampling Methods
Because studying an entire population is usually impractical, we study a sample. The sampling method affects the validity of conclusions dramatically. In simple random sampling, every member of the population has an equal probability of being selected. In stratified sampling, the population is divided into subgroups (strata) and random samples are drawn from each stratum, which improves representation. Systematic sampling selects every k-th member from a list. Cluster sampling divides the population into clusters and randomly selects entire clusters. Convenience sampling (choosing whoever is easily available) is prone to severe bias and should be avoided in serious research.
Bias in Data Collection
Bias is any systematic tendency to produce results that deviate from the truth. Common sources include: sampling bias (the sample does not represent the population, e.g., a voluntary online poll attracts only people with strong opinions — this is called voluntary response bias); non-response bias (certain types of people consistently refuse to respond, skewing results); and measurement bias (the measurement instrument itself consistently over- or under-records). Identifying and minimising bias is a central concern of study design. A study with large sample size n but systematic bias does not become more accurate — it becomes more precisely wrong.
The Key Equation
- \hat{p} — Sample proportion (dimensionless (0 to 1)): The proportion of the sample that has the characteristic being measured; used to estimate the true population proportion p.
- x — Number of successes (count): The count of individuals in the sample who possess the attribute of interest, such as agreeing with a statement in a survey.
- n — Sample size (count): The total number of individuals included in the sample. Larger n generally reduces sampling variability and improves the estimate of the population parameter.
Worked Example
Problem. A student wants to estimate the proportion of students in her school of 800 who walk to school. She surveys every 8th student on the school roll and receives responses from 97 out of 100 students contacted. Of the respondents, 34 walk to school. (a) What sampling method is used? (b) Calculate the sample proportion who walk to school. (c) Identify one possible source of bias.
Answer. Systematic sampling; sample proportion ≈ 0.351 (35.1%); possible non-response bias from the 3 students who did not respond.
Origins · 1 minFlorence Nightingale and how this was discovered
Florence Nightingale was a pioneering statistician and nurse who revolutionised data collection in medicine. During the Crimean War (1853–1856), she systematically collected mortality data from field hospitals and discovered that the majority of soldier deaths were caused by preventable infections (recorded in her data as 'Zymotic diseases') rather than battle wounds. She invented the polar area chart — sometimes called the 'coxcomb diagram' — to display her collected data visually and persuade government officials to improve sanitary conditions. Her rigorous approach to gathering and presenting data directly saved thousands of lives and established principles of evidence-based healthcare that remain central to statistics today. Nightingale was reportedly a gifted mathematician from childhood, and her mentor was the logician and mathematician James Joseph Sylvester. She was the first female member elected to the Royal Statistical Society in 1858.
Related Concepts
- Sampling Methods — Data collection relies on choosing an appropriate sampling method; understanding random, stratified, and systematic sampling is the direct next step after grasping data collection principles.
- Bias and Variability in Sampling — The validity of any data collection effort depends on identifying and minimising bias, making this concept a natural extension of understanding how collection methods affect conclusions.
- Experimental Design — Controlled experiments are one of the primary methods of data collection; study design principles such as randomisation, replication, and control groups directly govern what causal conclusions can be drawn.
- Descriptive Statistics — Once data is collected, the first analytical step is to summarise it using measures of centre and spread; descriptive statistics are the immediate application of collected data.