Chapter 9 — Use Of Statistical Tools

What is Statistics? Statistics is a branch of mathematics that deals with collection, presentation, analysis and interpretation of numerical data to make decisions in the presence of uncertainty. In economics it helps summarize economic phenomena, draw conclusions and support policy and business decisions.

Main Functions of Statistics

• Collection: Gathering relevant data (surveys, experiments, records).
• Presentation: Organising data into tables, charts and diagrams for clarity.
• Analysis: Summarising data with measures (mean, median, mode, dispersion).
• Interpretation: Drawing conclusions, making forecasts and supporting decisions.

Types of Data

• Qualitative (categorical): e.g., gender, occupation (nominal/ordinal).
• Quantitative (numerical): e.g., income, age (discrete or continuous).

Sources of Data

• Primary data: Collected firsthand (surveys, experiments).
• Secondary data: Collected earlier by others (official records, publications).

Population and Sample — Population is the full set of items/people under study; a sample is a subset chosen for study. Proper sampling methods ensure representativeness.

Steps in a Statistical Investigation

• Define objective and variables.
• Decide source and method of data collection.
• Collect data.
• Classify and tabulate data (frequency distribution).
• Summarise and analyse using measures and graphs.
• Interpret results and report findings.

Presentation of Data — Raw data are made meaningful by arranging into frequency distributions, tables, and by drawing diagrams and graphs: bar charts, histograms, pie charts, frequency polygons, ogives.

Measures Used — Two broad kinds: measures of central tendency (mean, median, mode) and measures of dispersion (range, variance, standard deviation, coefficient of variation). These describe the typical value and spread of data.

Uses of Statistics (especially in Economics)

• Summarise large amounts of economic data (national income, prices).
• Compare different groups or periods (growth rates, inflation).
• Forecast and inform policy (demand estimation, resource allocation).
• Test hypotheses (evaluate claims about populations).

Limitations and Precautions

• Quality depends on the quality of data; biased or incomplete data produce misleading results.
• Care needed in sample selection, question design, and interpretation.
• Correlation does not imply causation.

What is Collection of Data? Collection of data is the process of gathering information needed to answer a research question or to compile statistics. In economics and statistics it is the first and most important step in any empirical study because the quality of results depends on the quality of data collected.

Types of data by source:

• Primary data: Data collected first-hand for a specific purpose (e.g., through surveys, interviews, observation, experiments).
• Secondary data: Data already collected by others for some purpose (e.g., government publications, books, journals, administrative records, online databases).

Methods of collecting primary data

• Questionnaire / Schedule: Structured set of written questions. A questionnaire is usually filled by the respondent; a schedule is filled by the enumerator.
• Interview: Face-to-face, telephone, or online interviews—useful for detailed or qualitative responses.
• Observation: Recording behaviour/events directly (participant or non‑participant observation).
• Experimentation: Controlled collection of data by manipulating variables (used less often in basic economic surveys).

Sampling vs Census

• Census: Collecting data from every unit of the population — accurate but time-consuming and costly.
• Sample survey: Collecting data from a representative subset — less costly, quicker, with sampling error.

Important considerations and steps

1. Define objectives clearly (what to measure and why).
2. Define the population and sampling frame.
3. Choose a method (census or sample; decide on sampling technique: simple random, stratified, systematic, cluster, etc.).
4. Design instrument (questionnaire/schedule): simple language, clear order, pre-test (pilot survey).
5. Train enumerators and carry out data collection ethically (consent, confidentiality).
6. Validate, code, and record data carefully to avoid errors (checking for missing/illogical values).

Quality aspects: Reliability (consistency of measurement), validity (measuring what is intended), accuracy (closeness to true value), timeliness, and cost-effectiveness. Non-sampling errors (response bias, recording errors, non-response) are often more serious than sampling errors.

What is sampling? Sampling is the process of selecting a part (sample) of a population to draw inferences about the whole. Sampling is used when a census is impractical, costly, or time-consuming.

Key concepts: population, sampling frame (list from which sample is drawn), sample size, sampling unit, sampling error (difference between sample estimate and true population value), and bias (systematic error).

Major types of sampling

• Probability sampling — every unit has a known (non-zero) chance of selection. Common methods:
• Simple random sampling (SRS): every sample of size n from N has equal chance. Use random numbers or lotteries.
• Systematic sampling: choose every k-th unit from an ordered list after a random start (k = N/n).
• Stratified sampling: divide population into homogeneous strata (groups) and draw samples from each stratum. Improves precision when strata differ.
• Cluster sampling: divide population into clusters (usually naturally occurring groups like villages), randomly select clusters, and study all or a sample of units within chosen clusters.
• Multistage sampling: a generalization of cluster sampling where selection occurs in stages (for example, districts → schools → students).
• Probability proportional to size (PPS): clusters or units are selected with probability proportional to their size; useful when cluster sizes vary.
• Non-probability sampling — selection is not based on known probabilities. Common methods:
• Convenience sampling: choose easily available units (fast but high bias).
• Judgment/purposive sampling: researcher selects units judged to be typical or knowledgeable.
• Quota sampling: sample reflects certain characteristics of the population (age, sex) but selection within quotas is non-random.
• Snowball sampling: existing respondents recruit further respondents (useful for hard-to-reach populations).

How to choose a method? Use probability methods when you need unbiased estimates and the sampling frame exists. Use stratification to increase precision when subgroups differ. Use cluster or multistage sampling for wide geographic areas to reduce cost. Non-probability methods are suitable for exploratory or quick studies where inference to a larger population is not required.

Precision and sample size: precision improves with larger n. For many estimators the sample mean has an approximately normal sampling distribution (Central Limit Theorem) for moderately large n, enabling confidence intervals and hypothesis tests.

Advantages and disadvantages (brief)

• Probability sampling: less bias, allows estimation of sampling error; but requires frame and can be costlier.
• Non-probability sampling: cheaper and quicker; results are subject to selection bias and cannot support reliable measures of sampling error.

Common pitfalls: incomplete or biased sampling frame, nonresponse bias, improper use of systematic or quota selection without random start, treating non-probability samples as if they were probability samples.

What is Classification?
Classification is the process of arranging raw data into groups or classes based on common characteristics so that it becomes easier to study, compare and draw conclusions. It reduces complexity and reveals patterns.

Types of Classification

• By nature of data: Qualitative (categorical) vs Quantitative (numerical).
• By measurement: Discrete (countable) vs Continuous (measurable).
• By number of attributes: One-way (univariate), Two-way (bivariate) and Multi-way (multivariate) classification.

Important principles of good classification

• Mutually exclusive: an observation must belong to one class only.
• Exhaustive: all observations must be covered.
• Homogeneous within classes and heterogeneous between classes.
• Simple and convenient for analysis.
• Prefer equal class-widths for continuous data (unless there is reason not to).

What is Tabulation?
Tabulation is the systematic arrangement of classified data in rows and columns (tables) with clear headings. A good table has a title, labeled rows/columns, units, and totals where appropriate. Tabulation summarizes data and prepares it for graphical presentation and calculation of statistical measures.

Steps to classify and tabulate numerical data

1. Decide the objective and the variable(s) to be classified.
2. For quantitative continuous data determine range = Max – Min.
3. Choose number of classes (k) and class width (h).
4. Construct class intervals (mutually exclusive and exhaustive).
5. Assign each observation to a class and tally frequencies.
6. Compute relative frequency, percentage frequency and cumulative frequency.
7. Present results in a clear table with totals.

Simple example table (student marks)

Uses of classification & tabulation

• Simplify large datasets for interpretation.
• Prepare data for graphs (histogram, pie chart, ogive, etc.).
• Calculate central tendency, dispersion and other statistical measures.
• Compare subgroups (e.g., income classes, age groups, regions).

Diagrammatic and Graphic Presentation refers to methods of representing statistical data visually so patterns, trends and comparisons become easy to understand. Good diagrams/graphs make complex numerical information clear at a glance and are essential for communication in economics.

Key principles and conventions

• Every diagram must have a clear title, labeled axes (with units), an appropriate scale, and source/footnote if necessary.
• Choose the type of diagram that suits the data: qualitative categories → bar/column/pie; grouped frequency data → histogram, frequency polygon, ogive; paired numerical data → scatter diagram.
• Maintain correct proportions, avoid distorted scales, and include a legend when more than one series is shown.

Common types and how to construct them

• Bar (Column) Diagram: Represents categorical data by rectangular bars. Use gaps between bars for discrete categories. Bars can be simple, multiple (side-by-side for comparison), component (stacked parts of a total) or sub-divided.
• Pie Chart: Shows parts of a whole. Convert each category's share to an angle: angle = (frequency / total) × 360°. Use when categories sum to a meaningful total (e.g., budget composition).
• Histogram: For continuous grouped data. Draw contiguous rectangles whose heights equal class frequency (or frequency density if class widths are unequal). No gaps between bars.
• Frequency Polygon: Plot class mid-points on x-axis and join successive points by straight lines. Useful for comparing distributions.
• Ogive (Cumulative Frequency Curve): Plot cumulative frequency against upper (or lower) class boundaries; connects points to show cumulative distribution and to read medians/percentiles.
• Frequency Curve: A smoothed version of the frequency polygon; used to show general shape of distribution.
• Scatter Diagram: Plot paired observations (x,y) to study relationship or correlation. Add a line of best fit to indicate trend.

Advantages

• Quick visual comparison and pattern recognition (trends, peaks, gaps).
• Accessible to non-technical audiences.
• Highlights relationships between variables (with scatter diagrams) and composition (with pie/component bars).

Limitations

• Can be misleading if scales or proportions are manipulated.
• Lose precise numerical detail—diagrams summarize rather than show raw data.
• Some graphs (like pie charts) are unsuitable for many categories or very similar shares.

When to use which type (quick guide)

• Comparing categories: Bar/column diagram.
• Showing parts of a whole: Pie chart or stacked bar.
• Distribution of continuous data: Histogram, frequency polygon, frequency curve.
• Cumulative analysis (median, quartiles): Ogive.
• Studying relationship between two numerical variables: Scatter diagram.

Using these methods correctly helps students and analysts present economic data (like income distribution, expenditure composition, production trends, price movements) clearly and honestly.

Definition: Measures of central tendency are statistical values that describe the center or typical value of a dataset. The main measures are Arithmetic Mean, Median and Mode. They summarize data with a single representative number and are used to compare datasets, study distribution and make decisions.

Why they matter (Uses):

• Simplify large datasets to a single typical value (e.g., average income, average marks).
• Compare different groups (e.g., regions, years).
• Serve as inputs for other statistical measures (variance, standard deviation).

Desirable properties: They should be: representative, simple to compute, based on all observations (in case of mean), and stable (not too sensitive to small changes).

1. Arithmetic Mean (Average)
Definition: Sum of all observations divided by the number of observations. Useful when all values are equally important and data are not highly skewed or contain extreme outliers.

For ungrouped data (individual observations): x̄ = Σx / n

For grouped data (class intervals with mid-points x_i and frequencies f_i):
x̄ = (Σ f_i x_i) / N, where N = Σ f_i

Step-deviation method (useful for large class widths): choose a convenient origin a and class size h, set u_i = (x_i − a)/h, then
x̄ = a + h*(Σ f_i u_i / N)

2. Median
Definition: The middle value that divides the distribution into two equal parts (50% below and 50% above). Preferred for skewed distributions and when outliers are present.

For ungrouped (sorted) data:
- If n is odd → median is the ((n+1)/2)-th value.
- If n is even → median is the average of (n/2)-th and (n/2+1)-th values.

For grouped data (continuous classes): use linear interpolation within the median class:
Median = l + [((N/2) − c.f.) / f_m] * h

where l = lower boundary of median class, N = total frequency, c.f. = cumulative frequency before median class, f_m = frequency of median class, h = class width.

3. Mode
Definition: The value(s) that occur most frequently. Useful to identify the most common category or value (e.g., most common family size, most sold product).

For ungrouped data: the observation with highest frequency (can be unimodal, bimodal, multimodal).

For grouped data (continuous classes): estimate mode by interpolation using the modal class (class with maximum frequency):
Mode = l + [ (f1 − f0) / (2f1 − f0 − f2) ] * h

where l = lower boundary of modal class, f1 = frequency of modal class, f0 = frequency of class before, f2 = frequency of class after, h = class width.

Comparison & selection:

• Use mean when distribution is symmetric and data have no extreme outliers.
• Use median when distribution is skewed or contains outliers (it resists extremes).
• Use mode for categorical data or to find the most frequent value.

Limitations: Mean is sensitive to outliers; median ignores detailed values and only depends on rank; mode may be non-unique and unstable for small samples.

Practical tips for computation:

• Always check the data type: quantitative (mean, median, mode) vs categorical (mode only).
• For grouped data ensure class intervals are continuous and of equal width (or note the width in formulas).
• When using formulas for grouped data, use class mid-points for mean and linear interpolation for median and mode.

What are Measures of Dispersion?

Measures of dispersion describe how spread out or scattered the values of a data series are around a central value (mean, median or mode). While measures of central tendency (mean, median, mode) tell us the centre of data, measures of dispersion tell us the variability or consistency of the data. Common measures: Range, Quartile Deviation (QD), Mean Deviation (MD), Variance and Standard Deviation (SD).

Why they matter

Two distributions can have the same mean but very different spreads. Dispersion helps compare variability (e.g., marks of two classes, incomes of two cities, temperature changes), decide reliability of averages and assess risk (e.g., stock volatility).

Short descriptions

• Range: Difference between maximum and minimum values. Easiest but most affected by extremes.
• Quartile Deviation (QD) / Semi-interquartile range: Half the difference between third and first quartiles; resistant to outliers and shows spread of middle 50% of data.
• Mean Deviation (MD): Average of absolute deviations from a chosen central value (mean or median). MD about median gives minimum average absolute deviation.
• Variance: Average of squared deviations from the mean; gives squared units.
• Standard Deviation (SD): Square root of variance; same units as data and widely used because of good mathematical properties.
• Coefficient of Variation (CV): (SD/mean) × 100% — a relative measure useful to compare variability of series with different units or means.

Properties & Practical points

• All measures ≥ 0. Zero dispersion means all observations equal.
• Range and QD are simple and robust (QD more robust than range).
• Variance and SD use all observations and are sensitive to outliers (extreme values).
• SD has preferable algebraic properties (useful in statistical inference and probability).
• CV is dimensionless; use it to compare variability across datasets with different units or widely different means.
• For grouped data, use class mid-points (xi) and frequencies (fi).

How to compute (conceptual steps)

• Range: sort data, subtract min from max.
• QD: find Q1 (25th percentile) and Q3 (75th percentile) from sorted data (or from ogive for grouped data); QD = (Q3 - Q1)/2.
• MD: compute deviations from chosen center, take absolute values, average them (use frequencies for grouped data).
• SD: find mean, compute squared deviations, average (population) or average adjusted by (n-1) for sample, then square-root.

When to use which

• Use range or QD for quick, robust summary (QD when outliers present).
• Use MD when interest is in average absolute deviation (simpler interpretation).
• Use SD and variance for most statistical work, modelling and comparisons; use CV to compare across different units or scales.

Class 11 (CBSE) tips

• For grouped data, always use class mid-points to represent class values.
• When classes are open or unequal, take care in choosing representative values and class width h in computational shortcuts.
• Remember that median minimizes the sum of absolute deviations; mean minimizes the sum of squared deviations.

Definition: Correlation is a statistical measure that describes the degree and direction of association between two quantitative variables. In economics it shows how one economic variable changes when another changes (for example, income and consumption).

Types of correlation:

• Positive correlation: Both variables move in the same direction (e.g., study hours and marks).
• Negative correlation: Variables move in opposite directions (e.g., price and quantity demanded for a normal good).
• No correlation: No discernible relationship; points are scattered randomly.

Methods of studying correlation:

• Scatter diagram (graphical): Plot paired observations (x,y). The pattern gives a visual idea of direction and strength.
• Karl Pearson’s coefficient of correlation (r): Measures the degree of linear relationship between two variables measured on an interval/ratio scale.
• Spearman’s rank correlation (ρ): Used when data are ordinal or when we use ranks instead of raw values; useful for monotonic but not necessarily linear relationships.

Interpretation: The correlation coefficient (r or ρ) ranges from −1 to +1. Sign indicates direction (+ for positive, − for negative). Magnitude shows strength (values close to ±1 indicate strong association; close to 0 indicate weak or no linear association). Important caution: correlation does not imply causation. A high correlation can arise from coincidence, confounding variables, or indirect relationships.

Key properties (brief): Pearson's r is unitless, lies between −1 and +1, is unchanged by change of origin and scale (linear transformations), and measures only linear association; it is sensitive to outliers. Spearman’s ρ is less sensitive to outliers and measures monotonic association.

Use in economics: Correlation helps analyse relationships such as consumption vs income, advertising vs sales, education vs earnings, price vs demand, and to check multicollinearity among explanatory variables before regression analysis.

What is an Index Number?
An index number is a statistical measure that shows the relative change in a variable (or group of related variables) over time or between places, expressed with respect to a base value (usually 100). It simplifies complex data into a single figure that facilitates comparison.

Purpose & Uses

• Measure inflation (e.g., Consumer Price Index).
• Compare price or quantity changes over time or across regions.
• Adjust monetary figures for price changes (real vs nominal values).
• Summarize movements in baskets of goods (e.g., cost of living, wholesale prices).

Types of Index Numbers

• By variable: Price index (price changes), Quantity index (output or consumption changes), Value index (value change).
• By weighting: Simple (unweighted) index, Weighted index.
• By formula: Aggregate (ratio of sums), Price relatives (averages of item-wise ratios).

Common Methods / Formulas (conceptual)

• Simple Aggregate Price Index: (Sum of current prices / Sum of base year prices) × 100
• Weighted Aggregate Price Index (Laspeyres-type): (Sum of base year quantities × current prices) / (Sum of base year quantities × base prices) × 100
• Price Relatives: For each item, (Current price / Base price) × 100. Then average these relatives (simple or weighted).
• Laspeyres Index (L): uses base period weights (quantities).
• Paasche Index (P): uses current period weights.
• Fisher Ideal Index: geometric mean of Laspeyres and Paasche: sqrt(L × P).

Desirable Tests (Properties)

• Time reversal test: If you swap base and current periods, the reciprocal relation should hold.
• Factor reversal test: For price index P and quantity index Q, P × Q should equal the value index (if properly constructed).
• Circular test: Index from A→B→C should be consistent with A→C.

Step-by-step construction (typical Laspeyres price index)

1. Choose base year and current year.
2. Select a representative basket of goods and their base-year quantities.
3. Collect base-year prices and current-year prices for each item.
4. Compute weighted sum: Σ(q0 × p1) and Σ(q0 × p0).
5. Index = [Σ(q0 × p1) / Σ(q0 × p0)] × 100.

Short numerical example

Consider three goods A, B, C with base-year (0) quantities and prices and current-year (1) prices:

Item   q0   p0   p1
A      10   5    6
B      20   2    3
C      30   1    1.2

Compute Laspeyres price index:

Σ(q0×p1) = 10×6 + 20×3 + 30×1.2 = 60 + 60 + 36 = 156
Σ(q0×p0) = 10×5 + 20×2 + 30×1 = 50 + 40 + 30 = 120
Index = (156 / 120) × 100 = 130 → Prices up by 30% since base year

Interpretation
An index of 130 (base =100) means an average increase of 30% in the price level of the chosen basket since the base year. Choice of weights and items affects results—hence different indices (CPI, WPI, GDP deflator) can give different percentages.

Limitations

• Choice of base year and basket may bias results (substitution bias, quality changes).
• New goods and changes in quality require adjustments.
• Different formulae give different answers; no single perfect index.

Practical examples and uses
Governments and central banks use index numbers to measure inflation (CPI, WPI), to adjust wages, pensions, tax brackets, and to deflate nominal GDP to obtain real GDP.

What interpretation means
Interpretation of statistics is the process of drawing meaningful conclusions from numerical data. Good interpretation requires attention to context (what is measured, units, time period), the type of data (sample or population), measures used (mean, median, percentages), variability and reliability (sample size, standard error, significance) and the possibility of alternative explanations.

Principles for correct interpretation

• Context: Always ask who collected the data, when and why. Units and base years matter for indices and percentages.
• Understand the measure: Mean, median and mode can tell different stories; choose the one appropriate to the distribution.
• Look at variability: Two series with the same mean can have very different spreads—use variance/SD and range.
• Sample versus population: If data come from a sample, check sample size, sampling method and margin of error before generalising.
• Beware of outliers: Single extreme values can distort means and correlations.
• Time patterns: Distinguish trend from seasonal or cyclical fluctuations in time-series data.
• Correlation ≠ causation: A statistical association does not prove one variable causes the other—look for confounders, time order and plausible causal mechanism.
• Aggregation effects: Aggregating data (e.g., averages over regions) can hide important within-group differences (ecological fallacy).

Limitations of statistical methods

• Incomplete picture: Statistics describe measurable aspects; qualitative or unmeasured factors (preferences, culture, policy details) may be important.
• Sampling and measurement errors: Poor questionnaire design, non-response, misreporting and recording errors lead to wrong conclusions.
• Bias: Selection bias, non-random samples and survivorship bias produce misleading estimates.
• Time-lag and dynamics: Statistics are often backward looking and may not reflect rapid changes.
• Choice-dependence: Results can change with definitions, base year, or choice of index formula and averaging method.
• Over-simplification: Single summary statistics (like a mean) hide distributional issues such as inequality.

Common statistical errors (what to watch for)

• Confusing correlation with causation — assuming A causes B because they move together.
• Using an inappropriate average — using mean for highly skewed income data instead of median.
• Biased sampling — using a sample that is not representative (e.g., internet poll for national opinion without accounting for access differences).
• Misleading graphs — truncated axes, distorted scales, 3D effects, unclear labels or omitted baselines.
• Ignoring confounding variables — apparent effects disappear when a third variable is controlled for.
• Simpson’s paradox and ecological fallacy — aggregated data can reverse or hide relationships present in subgroups.
• Cherry-picking and data dredging — selective reporting of favorable results or searching until a significant result appears.
• Misuse of percentages and indices — failing to note base sizes or base-year effects (e.g., a 100% increase from 1 to 2 is small in absolute terms).

How to avoid mistakes

• Always check source, sample method and sample size.
• Compare mean, median and mode and examine dispersion (variance/SD, box-plot).
• Plot data—histograms, boxplots and scatter plots reveal structure that summary numbers hide.
• Test robustness—see if results hold under different definitions, subgroups and time periods.
• Look for alternative explanations and control for likely confounders where possible.

Summary
Good interpretation combines numerical computation with judgement about data quality, context and plausible causal stories. Awareness of limitations and common errors helps avoid misleading conclusions.