Collection, Organisation & Presentation of Data Notes

The first decision in any study is the source of the data. Primary data are collected afresh by the investigator for the immediate purpose — a questionnaire filled by households in a survey. Secondary data are figures that already exist, collected earlier by someone else for another purpose and now reused, such as a student using the Census of India for a project.

• Primary data: original, first-hand, collected for the purpose in hand, by questionnaire, interview, observation or experiment.
• Secondary data: previously collected, published and re-analysed — the Census, government digests, RBI bulletins, the reports of the NSSO and CSO.
• When to prefer primary: the available secondary figures are inadequate, outdated or collected for a different objective.
• When to prefer secondary: time and cost are scarce, and existing data serve the purpose with suitable adjustment.

Once the investigator decides to collect primary data, the next question is whether to count everyone or to count a part. The census enumerates every unit; the sample enumerates a selected part and generalises. The sample must be chosen so that every unit of the population has an equal chance of being included — that is what makes it representative.

• Census: complete enumeration, highest accuracy, very costly and time-consuming, limited in scope by money and manpower.
• Sample: a part of the population examined under the same conditions; the results are then inflated to the whole.
• Advantages of sampling: economy in money and time, manageable data, higher accuracy because the volume is small enough to verify, and feasible where a census is impossible, as in quality testing of bulbs.
• Disadvantages: risk of a non-representative sample, and sampling error in every estimate.
• Random sampling methods: simple random sampling (lottery), stratified random sampling (population divided into groups then sampled from each), systematic sampling (every k-th unit), and cluster sampling (groups chosen at random).
• Non-random (purposive) methods: convenience sampling, judgement sampling and quota sampling, chosen for speed but weaker on representativeness.

India's secondary data are produced by a small set of central organisations, and naming them with their function is a reliable short-answer question.

• NSSO, the National Sample Survey Office: the largest survey agency, conducting nationwide socio-economic sample surveys on consumer expenditure, employment and industry.
• CSO, the Central Statistical Office: compiles national income, the Index of Industrial Production and statistical abstracts.
• Census of India: the population census held every ten years by the Office of the Registrar General.
• RBI: publishes money, banking and balance-of-payments statistics.
• Ministries and departments: agriculture, industry and commerce ministries publish their own statistical abstracts.
• International agencies: the World Bank, the IMF and the UN yearbooks are standard sources for comparisons.

Raw data are a mass of unclassified figures. Organisation classifies them into a frequency distribution in which each value or group of values is paired with the number of times it occurs. The variable may be discrete, taking isolated values such as 0, 1, 2, or continuous, taking every value over a range such as income between 10 and 20.

• Discrete variable: countable separate values — number of children, number of workers.
• Continuous variable: an unbroken range — income, height, weight, price.
• Class interval: the group of values, such as 10–20.
• Class limits: the lower limit 10 and the upper limit 20.
• Class size (width): upper limit − lower limit, so 20 − 10 = 10.
• Mid-value (mid-point): (lower limit + upper limit) ÷ 2, the representative of the class.
• Frequency: the number of observations falling in the class.
• Tally marks: strokes counted in groups of five to build the frequency.

The first form of presentation is the table, a systematic arrangement of data in rows and columns. A statistical table carries fixed parts, and the examiner may ask for the parts or ask for a table to be drawn from a set of figures.

• Table number: identifies the table in the study.
• Title: what the table shows, its subject, source and time.
• Caption: the column headings — what each column measures.
• Stubs: the row headings — the categories listed in the leading column.
• Sub-headings: subdivisions of the caption and stubs.
• Body: the actual figures.
• Headnote: the unit of measurement, such as 'in crore rupees'.
• Source note: where the data came from — a requirement of honesty.
• Footnote: the special explanations.

Diagrams carry data in picture form. Bars suit comparisons among categories; the pie suits the share of each component in the total. Both are drawn only for discrete or one-variable data, and both appear in the paper as identify, draw or interpret questions.

• Simple bar diagram: one bar per category, height proportional to the value.
• Multiple bar diagram: a group of bars for each category, one bar per component, for comparing components.
• Subdivided (component) bar diagram: one bar per category split into segments in proportion to the components, best for showing shares.
• Pie diagram: a circle divided into sectors, each sector's angle proportional to the component's share of the total.
• Rule for the pie: angle of the sector = (component value ÷ total) × 360°.
• Beware: a pie suits shares, a multiple bar suits comparisons of magnitude — do not interchange them in the interpretation.

Line graphs plot continuous variables against a scale. Four curves dominate this unit and each answers a different question about the data: the histogram for class frequencies, the frequency polygon for comparison, the cumulative ('ogive') curves for medians, and the time series line graph for change over time.

• Histogram: rectangles on a continuous scale, area proportional to the frequency — for a continuous frequency distribution.
• Frequency polygon: the class mid-points plotted and joined by straight lines; can be drawn over the histogram by joining the tops of the rectangles.
• Ogive (cumulative frequency curve): cumulative frequencies plotted against the upper (less-than) or lower (more-than) limits, used to read the median and other values.
• Time series line graph: values plotted against time, such as import figures year by year, to show trend.
• Choosing the tool: histogram for classes, polygon for two series on one chart, ogive for medians, line graph for trends.