Business Statistics and Research Methods — UGC NET Commerce (Paper 2)

Statistics chapters have a reputation for being all formula and no substance, but UGC NET treats this one as a numeracy check disguised as a theory paper — a short data set, a formula you're expected to already know, and four numerically close options waiting to catch whichever arithmetic slip you make along the way. The fastest route through this chapter is not more definitions; it's re-deriving each formula on real numbers until the steps stop feeling unfamiliar.

1. What UGC NET actually asks

This chapter carries roughly 11% of Commerce Paper 2 — about 11 of the paper's 100 questions, worth 22 of the paper's 200 marks at +2 each, with no negative marking anywhere on the paper. That marking scheme has a direct tactical consequence for this chapter in particular: a numerical statistics question that looks intimidating at first glance is still worth attempting once you've narrowed it to two options, because an educated guess costs nothing that leaving it blank wouldn't have cost anyway.

Four broad question types recur here:

1. Direct computation — given a small data set, compute the mean, median, mode, variance, standard deviation, or correlation coefficient.
2. Term and method identification — matching a described sampling procedure or research step to its correct name.
3. Scenario-based inference concepts — identifying a Type I or Type II error, or correctly stating a null versus alternative hypothesis, from a described situation.
4. Sequencing — ordering the stages of the business research process correctly.

Because the syllabus spans both computational statistics and the more conceptual vocabulary of research methodology, candidates who prepare only one half tend to leave easy marks behind on the other. This chapter builds both halves side by side, with every statistical formula demonstrated on an actual worked data set.

2. Measures of central tendency

Central tendency answers the question "what is a typical value in this data set?" — and the three standard answers each capture something slightly different.

Arithmetic mean — the sum of all observations divided by their count: Mean = (Sum of all values) ÷ (Number of values). It uses every observation but is sensitive to extreme values (outliers can pull it noticeably away from where most of the data actually sits).

Median — the middle value once the data is arranged in ascending or descending order. For an odd number of observations, it is the single middle value; for an even number, it is the average of the two middle values. It is far less sensitive to extreme values than the mean, which is exactly why income and property-price data are often reported using medians rather than means.

Mode — the value that occurs most frequently in the data set. A data set can have one mode (unimodal), more than one mode (bimodal or multimodal), or, if every value is equally frequent, no mode at all.

Worked example

Consider five days of a shop's sales figures (in ₹'000): 10, 20, 20, 40, 50.

Mean = (10 + 20 + 20 + 40 + 50) ÷ 5 = 140 ÷ 5 = 28.

Median — the data is already sorted; with 5 (odd) observations, the median is the 3rd value = 20.

Mode — 20 appears twice, more often than any other value, so the mode is 20.

Notice the mean (28) sits well above both the median and the mode here — a signature of a data set pulled upward by one large value (₹50,000), a pattern worth recognising quickly in any scenario question about skewed data.

3. Measures of dispersion

Central tendency alone can be misleading — two data sets can share an identical mean while one is tightly clustered and the other wildly spread out. Dispersion measures quantify exactly that spread.

Range — the simplest measure: the difference between the highest and lowest values. Crude, since it uses only two observations and ignores everything in between, but quick to compute.

Variance — the average of the squared deviations of each observation from the mean: Variance = (Sum of squared deviations from the mean) ÷ N, using N for a population and (N − 1) for a sample. Squaring the deviations serves two purposes — it prevents positive and negative deviations from cancelling each other out, and it penalises larger deviations disproportionately.

Standard deviation (SD) — simply the positive square root of the variance, which brings the measure back into the same units as the original data (variance is in squared units, which aren't directly interpretable).

Coefficient of variation (CV) — standard deviation expressed as a percentage of the mean: CV = (SD ÷ Mean) × 100. Because it's a relative measure, CV is the correct tool whenever you need to compare the variability of two data sets measured in different units or on very different scales — comparing raw standard deviations directly in that situation is a common and costly mistake.

Worked example — continuing the same data set

Using the same five sales figures (10, 20, 20, 40, 50), with mean 28:

Sum of squared deviations = 324 + 64 + 64 + 144 + 484 = 1,080.

Variance (population) = 1,080 ÷ 5 = 216.

Standard deviation = √216 ≈ 14.70.

Range = 50 − 10 = 40.

Coefficient of variation = (14.70 ÷ 28) × 100 ≈ 52.5% — a notably high figure, confirming visually what the wide range already suggested: this is a highly dispersed data set relative to its own mean.

4. Correlation and regression

Correlation measures the strength and direction of the linear relationship between two variables, without implying that one causes the other. Karl Pearson's coefficient of correlation (r) is the standard measure, computed from paired data (X, Y) as:

r = [n·ΣXY − ΣX·ΣY] ÷ √{[n·ΣX² − (ΣX)²] × [n·ΣY² − (ΣY)²]}

r always falls between −1 and +1: values near +1 indicate a strong positive relationship, values near −1 a strong negative relationship, and values near 0 indicate little to no linear relationship. Regression goes a step further, fitting an actual predictive equation — the regression line of Y on X, Y = a + bX — where b is the regression coefficient (the estimated change in Y for a one-unit change in X) and a is the intercept.

Worked example

A firm records its advertising expenditure (X, ₹ lakh) and sales (Y, ₹ lakh) across four months: X = 2, 4, 6, 8 and Y = 3, 7, 5, 9.

With n = 4: numerator = (4 × 136) − (20 × 24) = 544 − 480 = 64. First bracket = (4 × 120) − 20² = 480 − 400 = 80. Second bracket = (4 × 164) − 24² = 656 − 576 = 80. Denominator = √(80 × 80) = 80.

r = 64 ÷ 80 = 0.8 — a strong positive correlation between advertising spend and sales.

The regression coefficient of Y on X uses the same numerator and first bracket: b = 64 ÷ 80 = 0.8. The intercept a = (ΣY − b·ΣX) ÷ n = (24 − 0.8 × 20) ÷ 4 = (24 − 16) ÷ 4 = 2. The regression line of Y on X is therefore Y = 2 + 0.8X — for every additional ₹1 lakh spent on advertising, sales are predicted to rise by ₹0.8 lakh.

5. Probability basics

Classical probability defines the probability of an event as the number of favourable outcomes divided by the total number of equally likely outcomes: P(A) = (Favourable outcomes) ÷ (Total outcomes). Two theorems govern how probabilities combine:

• Addition theorem — for two events A and B, P(A or B) = P(A) + P(B) − P(A and B), where the last term corrects for double-counting any outcome common to both events.
• Multiplication theorem — for two independent events, P(A and B) = P(A) × P(B); for dependent events, this uses conditional probability instead: P(A and B) = P(A) × P(B | A).

Conditional probability, P(A | B), is the probability of A occurring given that B has already occurred, computed as P(A and B) ÷ P(B).

Worked example — a card is drawn at random from a well-shuffled deck of 52 playing cards. What is the probability it is either a king or a spade? P(King) = 4/52, P(Spade) = 13/52, and P(King of Spades, the overlap) = 1/52. By the addition theorem: P(King or Spade) = 4/52 + 13/52 − 1/52 = 16/52 (= 4/13 ≈ 0.308).

6. Sampling methods

A sample is a subset of a population studied to draw conclusions about the whole population without surveying every unit. Sampling methods split into two families:

The core distinction UGC NET tests repeatedly: probability sampling gives every unit a known, non-zero chance of selection and supports statistical inference about the wider population; non-probability sampling does not, and is used mainly for exploratory or hard-to-access research where a strict probability sample isn't feasible.

7. Hypothesis testing and the research process

Hypothesis testing formalises how a business researcher decides whether a pattern observed in sample data reflects a real effect or is plausibly just chance variation. The null hypothesis (H0) is a statement of no difference or no effect — the status quo — while the alternative hypothesis (H1) is the statement the researcher is actually trying to find evidence for. A test either rejects H0 in favour of H1, or fails to reject H0; it never "proves" H0 true.

Two kinds of error are possible, and the courtroom analogy makes both memorable: if the null hypothesis is "the defendant is innocent," then convicting an innocent defendant is a Type I error (rejecting a null hypothesis that is actually true, with probability denoted α, the level of significance), while acquitting a genuinely guilty defendant is a Type II error (failing to reject a null hypothesis that is actually false, with probability denoted β). Lowering α (making it harder to convict/reject) generally raises the risk of β, and vice versa — the two error types trade off against each other for a fixed sample size.

The business research process itself typically runs through a fixed sequence of stages: (1) formulating the research problem (defining precisely what question is being investigated), (2) reviewing existing literature, (3) developing the research design (deciding whether the study will be exploratory, descriptive, or causal in nature), (4) determining the sampling design, (5) collecting data (via primary tools like questionnaires, interviews, and observation, or secondary sources), (6) analysing and interpreting the data, and (7) preparing and presenting the research report. Note that the problem must be formulated before the design or the sampling plan can be sensibly chosen — a sequencing detail UGC NET tests directly.

8. Solved PYQ-style examples

Q1. A batsman's scores across 7 innings are: 25, 30, 30, 45, 50, 55, 60. Find the mean, median, and mode of this data. Solution. Mean = (25+30+30+45+50+55+60) ÷ 7 = 295 ÷ 7 ≈ 42.14. Median (4th value of 7, sorted) = 45. Mode = 30 (the only value appearing more than once). Answer: Mean ≈ 42.14, Median = 45, Mode = 30.

Q2. For the data set 4, 6, 8, 10, 12 (treated as the full population), compute the variance and standard deviation. Solution. Mean = 40 ÷ 5 = 8. Deviations: −4, −2, 0, 2, 4; squared: 16, 4, 0, 4, 16; sum = 40. Variance = 40 ÷ 5 = 8. Standard deviation = √8 ≈ 2.83. Answer: Variance = 8, Standard deviation ≈ 2.83.

Q3. For the paired data X: 1, 2, 3, 4 and Y: 2, 3, 5, 4, compute Karl Pearson's coefficient of correlation. Solution. ΣX=10, ΣY=14, ΣXY=39, ΣX²=30, ΣY²=54, n=4. Numerator = (4×39) − (10×14) = 156 − 140 = 16. First bracket = (4×30) − 100 = 20. Second bracket = (4×54) − 196 = 20. Denominator = √(20×20) = 20. r = 16 ÷ 20 = 0.8. Answer: r = 0.8 (a strong positive correlation).

Q4. A researcher surveying small retail shop owners across a state divides the state into districts, randomly selects a handful of districts, and then surveys every shop owner within those selected districts, rather than sampling individually across the whole state. Which sampling method is this? Solution. Entire natural groups (districts) are randomly selected, with every unit inside a chosen group included — this is the defining feature of cluster sampling, distinct from stratified sampling, which samples proportionately within every stratum rather than selecting only some clusters wholesale. Answer: Cluster sampling.

Q5. A batch of electronic components that actually meets the required quality specification is mistakenly rejected by an automated inspection test. What kind of statistical error does this illustrate? Solution. If H0 is "the batch meets the quality standard," rejecting this true null hypothesis is precisely the definition of a Type I error. Answer: Type I error.

Q6. A company wants to test whether a new sales-training program has changed average monthly sales per employee, currently ₹80,000. State the correctly formulated null and alternative hypotheses. Solution. The null hypothesis always represents "no change" from the known or assumed value, and the alternative represents the change being investigated. Answer: H0: population mean monthly sales = ₹80,000 (unchanged); H1: population mean monthly sales ≠ ₹80,000 (changed).

Q7. Salesperson A has mean monthly sales of ₹50,000 with a standard deviation of ₹5,000. Salesperson B has mean monthly sales of ₹80,000 with a standard deviation of ₹6,000. Despite B's larger absolute standard deviation, which salesperson shows greater RELATIVE variability in sales? Solution. CV(A) = (5,000 ÷ 50,000) × 100 = 10%. CV(B) = (6,000 ÷ 80,000) × 100 = 7.5%. Comparing relative, not absolute, dispersion reverses the naive conclusion drawn from the standard deviations alone. Answer: Salesperson A (CV = 10%, versus B's CV = 7.5%).

Q8. Arrange the following stages of the business research process into their correct order: (i) Data collection (ii) Formulating the research problem (iii) Report writing (iv) Data analysis and interpretation (v) Research design. Solution. The process must begin with defining the problem, since every later stage — literature review, design, sampling, collection — depends on knowing exactly what question is being investigated. Answer: (ii) Formulating the research problem → (v) Research design → (i) Data collection → (iv) Data analysis and interpretation → (iii) Report writing.

9. Common traps

• Confusing variance with standard deviation — variance is in squared units and is often mistaken for the final answer when the question actually asks for standard deviation (its square root), or vice versa.
• Using the wrong denominator for variance — population variance divides by N; sample variance divides by (N − 1); mixing these up produces a subtly wrong figure that can still look plausible.
• Comparing raw standard deviations across data sets measured in different units or scales — this is exactly what the coefficient of variation exists to fix; a larger absolute SD does not always mean greater relative variability.
• Treating correlation as proof of causation — a strong r only establishes a linear association between two variables, never that one causes the other.
• Mixing up probability and non-probability sampling methods — cluster sampling selects whole groups; stratified sampling samples proportionately from within every group; neither is the same as the non-random convenience or quota methods.
• Reversing Type I and Type II errors — Type I is rejecting a null hypothesis that is actually true; Type II is failing to reject a null hypothesis that is actually false; the courtroom analogy (wrongful conviction vs. wrongful acquittal) is the fastest way to keep these straight under pressure.
• Misstating the null hypothesis — H0 always represents no difference or no effect (the status quo being tested), never the change or effect the researcher hopes to demonstrate; that role belongs to H1.
• Sequencing the research process incorrectly — the research problem must be formulated before the research design or sampling plan can be chosen; data collection can never sensibly precede problem formulation.

10. Training protocol

Build fluency in this chapter by working every core formula — mean, median, mode, variance, standard deviation, Karl Pearson's r, and the regression line — on at least one small data set of your own choosing until you can move from raw numbers to a final answer without hesitating over which sum goes in which part of the formula; the exam's numerical questions are engineered to punish exactly that hesitation with four closely spaced wrong options. Keep a strict mental separation between the computational half of this chapter (central tendency, dispersion, correlation, probability) and the conceptual half (sampling methods, hypothesis-testing vocabulary, the research process), since UGC NET draws roughly evenly from both, and over-preparing one at the expense of the other is the single most common revision mistake for this chapter. Anchor Type I and Type II errors permanently to the courtroom analogy, and anchor the seven-stage research process to the simple logic that you cannot design a study, sample a population, or collect data before you have actually stated what question you're trying to answer.