Chapter 6 — Measures Of Dispersion

Meaning: Dispersion (or variability) describes how spread out or scattered the values of a data set are around a central value (like mean or median). While averages (mean, median, mode) give a central tendency, dispersion tells us whether data points are clustered tightly around that center or are widely spread. Common measures of dispersion include range, interquartile range (IQR), mean deviation, variance and standard deviation, and coefficient of variation.

Types of measures: Absolute measures (range, IQR, mean deviation, variance, standard deviation) give spread in original units. Relative measures (coefficient of variation) express spread relative to the mean, allowing comparison between different data sets.

Objectives / Why we measure dispersion:

• To assess reliability of an average: small dispersion means the average represents the data well; large dispersion means the average may be misleading.
• To compare variability of two or more distributions even when they have different units or means (use relative measures like coefficient of variation).
• To understand stability: e.g., stable prices, incomes or production show low dispersion over time; volatile ones show high dispersion.
• To identify outliers and the spread of middle 50% (using IQR and box plots).
• To inform decisions in business, policy and risk management — e.g., investment risk (higher dispersion of returns = higher risk), wage inequality, exam performance variability.
• To choose appropriate statistical methods: some tests assume low/high variability or require measures of dispersion as inputs.

Interpretation: Two distributions can have the same mean but very different dispersions — measuring dispersion helps detect such differences. Relative measures (CV) are useful when comparing data with different units or scales. Measures like IQR and box-plot visualization help spot skewness and outliers, while variance/standard deviation are used in many further analyses (e.g., probability, sampling).

A measure of dispersion describes how stretched or compressed the values of a variable are around a central value (like the mean or median). A good measure of dispersion should satisfy certain qualities so it is useful, interpretable and mathematically reliable. Key qualities are:

• Based on all observations: It should use information from every observation so that no data are ignored. Measures based on all observations (e.g., variance, standard deviation, mean deviation) reflect the entire distribution; measures like the simple range use only two values and may be misleading.
• Simple and easy to understand: The concept and computation should be reasonably straightforward for users. Simple measures (range, interquartile range, mean deviation) are easier to explain than higher-order measures, though trade-offs exist.
• Definite and unique (rigidly defined): The rule/formula must be unambiguous and give a single value for any data set so results are reproducible.
• Finite and defined for all data sets: The measure should give a finite number for practical data. Measures that can diverge or are undefined in normal situations are not useful.
• Not affected by change of origin: If every observation is increased or decreased by the same constant (x_i -> x_i + c), the measure of dispersion should not change. This ensures location shifts do not alter spread. (Variance, standard deviation, MD, range satisfy this property.)
• Proper behaviour under change of scale: If all observations are multiplied by a constant a (x_i -> a x_i), the measure should change in a definite way: scale it by |a| (for SD, MD, range) or by a^2 (for variance). This makes comparisons under unit changes meaningful.
• Not unduly affected by sampling fluctuations: A good measure should be reasonably stable when samples change slightly (robustness). Some measures (e.g., median-based IQR) are more robust to outliers than variance/SD.
• Amenable to algebraic and statistical treatment: It should allow further mathematical manipulation and inference (e.g., variance can be used in hypothesis tests and in properties of sums of variables).

In practice no single measure is perfect for every situation. Choice depends on whether you prioritise simplicity, sensitivity to outliers, or mathematical tractability.

Overview: Measures of dispersion describe how spread out or scattered a set of observations is. In Class 11 Economics the measures are classified into two broad types: Absolute measures (give dispersion in original units) and Relative measures (express dispersion relative to a central value, usually as a ratio or percentage).

• Absolute measures — common ones taught: Range, Quartile Deviation (QD), Mean Deviation (MD) or Mean Absolute Deviation, Variance and Standard Deviation (SD). These quantify spread in the same units as the data (except variance, which is in squared units).
• Relative measures — include coefficients formed from absolute measures so comparisons across different units or scales are possible: Coefficient of Range, Coefficient of Quartile Deviation, Coefficient of Mean Deviation, and Coefficient of Variation (CV).

Details of main measures:

• Range: Simplest measure. Range = maximum value − minimum value. Easy to compute but highly sensitive to extreme values (outliers).
• Quartile Deviation (QD) / Semi-interquartile range: Uses the middle 50% of data so it is resistant to extremes.
  • QD = (Q3 − Q1) / 2, where Q1 and Q3 are first and third quartiles.
• Mean Deviation (MD): Average of absolute deviations from a central value (mean, median or mode). Often mean or median is used as the centre.
  • MD = (Σ |xi − center|) / N (for ungrouped population data). For grouped data use class midpoints and frequencies.
  • MD is easy to understand (average deviation) but algebraically less tractable than variance/SD.
• Variance and Standard Deviation (SD): Most widely used. Variance = average squared deviation; SD = square root of variance.
  • Population variance: σ² = Σ(xi − μ)² / N. Population SD: σ = sqrt(σ²).
  • Sample variance (introduced later/used when working with samples): s² = Σ(xi − x̄)² / (n − 1) (Bessel's correction); s = sqrt(s²).
  • SD is algebraically convenient and appears in many statistical results (e.g., normal distribution).
• Relative measures (coefficients):
  • Coefficient of Variation (CV) = (SD / mean) × 100% — useful to compare variability of different series with different units or means.
  • Coefficient of Quartile Deviation = (Q3 − Q1) / (Q3 + Q1).
  • Coefficient of Mean Deviation = (MD / mean) × 100%.

Key properties and practical notes:

• All measures ≥ 0; they are 0 only when all observations are equal.
• Adding a constant to every observation does not change any measure of dispersion. Multiplying all observations by a constant k multiplies dispersion by |k| (and variance by k²).
• Range is easiest but least reliable when outliers exist. QD and MD are more robust. SD is most commonly used for inferential work and modeling.
• Use relative measures (especially CV) when comparing dispersion across series with different units or different means (e.g., incomes vs. prices).

Definition: Range is the simplest measure of dispersion. It is the difference between the largest and the smallest observation in a data set and shows the spread of values.

Formula (basic): Range = Maximum value − Minimum value

Interpretation: A larger range means greater spread in the data. Range gives a quick sense of variability but does not use all observations (only extremes) and is sensitive to outliers.

Computation for different types of data:

• Ungrouped (raw) data: Identify the maximum and minimum values and subtract. Example: for marks {45, 50, 60, 70, 90}, Range = 90 − 45 = 45.
• Grouped data (continuous classes): Use class (real) boundaries: Range = Upper boundary of highest class − Lower boundary of lowest class. Example: if first class is 10–19 and last class is 70–79, use real limits 9.5 and 79.5 so Range = 79.5 − 9.5 = 70.
• Grouped data (discrete class limits): If classes are inclusive integers (e.g., 10–19), you should convert to real limits (lower limit − 0.5, upper limit + 0.5) before computing range.

Relative/Standardised forms: To compare dispersion across different units or scales, use relative measures based on range:

• Relative Range (%) = (Range / Mean) × 100. Useful to compare variability of two series with different means.
• Coefficient of Range = (Max − Min) / (Max + Min). This is dimensionless and lies between 0 and 1 for positive data.

Properties, advantages and limitations:

• Advantage: Simple to compute and easily understood.
• Limitation: Uses only two observations (min and max) so ignores information in the rest of the data.
• Limitation: Highly sensitive to outliers/extreme values — one extreme value can greatly increase the range.
• Usefulness: Good for quick, rough comparisons; not suitable when a robust measure of dispersion is required (use variance, standard deviation, or interquartile range instead).

How teachers/students should present it (Class 11 level): Show worked numerical examples for both raw and grouped data, demonstrate effect of an outlier, and compare range with interquartile range or standard deviation to explain limitations.

Definition: The Interquartile Range (IQR) is the range of the middle 50% of a data set and is defined as IQR = Q3 − Q1, where Q1 (first quartile) is the value below which 25% of observations lie and Q3 (third quartile) is the value below which 75% of observations lie. The Quartile Deviation (QD), also called semi-interquartile range, is half the IQR: QD = (Q3 − Q1)/2.

Why it is used: IQR/QD measure the spread of the central half of the data and are robust (not sensitive to extreme values/outliers). They are preferred when the median is used as a measure of central tendency.

How to find quartiles (brief steps):

• For ungrouped (individual) data: sort the data in ascending order, find positions of quartiles using (n+1)/4 for Q1 and 3(n+1)/4 for Q3. If the position is not an integer, interpolate (take weighted average of neighbouring observations).
• For grouped (frequency) data: use cumulative frequencies to find the quartile class (where cumulative frequency first equals or exceeds n/4 or 3n/4), then interpolate inside that class using the class width.

Properties:

• IQR ignores extreme values and gives a measure of central spread.
• IQR ≥ 0; IQR = 0 when at least 50% of observations have the same value.
• QD = IQR/2 is used when a scale comparable to standard deviation or mean deviation is needed but robust against outliers.

Comparison with other measures: Unlike range, IQR is not affected by extremes. It complements the median (central tendency) in describing distribution.

Definition: Mean Deviation (also called Mean Absolute Deviation, MAD) is the average of the absolute differences between each observation and a central value (usually the mean or the median). It measures how spread out data values are around a central value.

Interpretation: A larger mean deviation means observations are more spread out; a smaller value means they are clustered near the central value. MD is expressed in the same units as the original data.

Types (by choice of central value):

• Mean deviation about mean: average of |x_i - x̄|.
• Mean deviation about median: average of |x_i - M| (often used when distribution is skewed).

Why use MD? It is easy to understand and less affected by extreme values than variance/standard deviation because it uses absolute values, not squares. However, it is not algebraically convenient for further mathematical manipulation.

General steps to compute MD:

• Choose the central value (mean or median).
• Find each deviation: d_i = x_i - (central value).
• Take absolute values: |d_i|.
• Compute average of these absolute deviations.

For grouped data: replace each observation x_i by class mark (midpoint) and multiply absolute deviations by the class frequency, then divide by total frequency.

Properties & Notes:

• MD >= 0. MD = 0 only when all observations are equal.
• Units of MD are same as data units.
• Less sensitive to outliers than variance/SD, but it does not have nice algebraic properties (e.g., not easily decomposable).

Definition and purpose
Variance and standard deviation are measures of dispersion that quantify how far observations in a data set are spread out from the mean. Variance measures the average squared deviation from the mean; standard deviation is the square root of variance and is in the same units as the original data.

Why both are used
Variance is useful in algebraic and theoretical work (adds for independent components), but its units are squared. Standard deviation (SD) is easier to interpret because it has the same units as the data and gives a typical distance from the mean.

Basic idea (ungrouped data)
1. Compute the mean (average) μ or x̄.
2. For each observation x, compute deviation (x − mean).
3. Square deviations and average them → variance.
4. Take square root → standard deviation.

Population versus sample
- Population variance (σ²) uses denominator N (the population size).
- Sample variance (s²) uses denominator (n − 1) (Bessel's correction) to provide an unbiased estimator of population variance from a sample.

Grouped data
For frequency (grouped) data, use class mid-points (x) and frequencies (f). Compute Σf, Σfx and Σfx², then apply the formula for variance. Shortcut methods (assumed mean or coding) simplify arithmetic.

Important properties

• Variance and SD are always ≥ 0.
• Variance is minimized when deviations are taken about the mean.
• SD is sensitive to outliers (large deviations increase SD a lot).
• Units: variance in (original units)²; SD in original units.
• Coefficient of variation (CV = SD/mean × 100%) allows comparison of relative variability across series with different units or means.

Interpretation
Approximately, for roughly symmetric/normal data: about 68% of observations lie within ±1 SD of the mean, about 95% within ±2 SD, and about 99.7% within ±3 SD (empirical rule).

When to use
Use variance and SD when you need a measure sensitive to all deviations (not just median or quartile spread). They are widely used in economics (income variability, risk/volatility, quality control).

What it is: The Coefficient of Variation (CV) is a relative measure of dispersion that shows the size of the standard deviation in relation to the mean of the distribution. It is dimensionless and usually expressed as a percentage. CV helps compare variability of two or more series that have different units or very different means.

Formula and interpretation: CV = (Standard Deviation / Mean) × 100%. A lower CV indicates less relative variability (more consistency) and a higher CV indicates greater relative variability. CV is useful when comparing risk or variability across datasets with different scales (for example, comparing returns of two investments with different average returns).

How to compute (steps): 1) Calculate the mean (μ or x̄). 2) Calculate variance and then the standard deviation (σ or s). 3) Divide SD by the mean and multiply by 100 to get CV as a percentage.

Limitations and cautions: If the mean is zero CV is undefined. If the mean is negative, CV may be negative — practitioners often use the absolute value of the mean for interpretation. CV is not appropriate when the mean is near zero or when the sign of the mean is meaningful in context. Also CV uses the mean and SD, so it is sensitive to outliers.

Why it matters in Economics (Class 11 context): CV is widely used to compare variability of incomes, prices, production yields, test scores, or returns across regions, time periods or investment choices where raw SDs are not comparable because the averages differ.

Definition: A change of origin means adding (or subtracting) a constant a to every observation: Y = X + a. A change of scale means multiplying every observation by a constant b: Y = bX. Combined: Y = bX + a.

What happens to measures of central tendency and dispersion?

- Mean: If X has mean μ_X then for Y = bX + a, mean μ_Y = b·μ_X + a. In particular, for a change of origin (b = 1) μ shifts by a; for scale only (a = 0) μ is multiplied by b.

- Deviations from the mean: (x_i + a) − (μ_X + a) = x_i − μ_X, so adding a constant (change of origin) does NOT change deviations about the mean. Multiplying by b multiplies each deviation by b.

- Range and mean deviation (about mean): Range is unaffected by addition and is multiplied by |b| when scaled. Mean deviation about mean is invariant under origin change and scales by |b| under scale change.

- Variance and standard deviation: Variance is invariant under addition and multiplied by b^2 under scaling. Standard deviation is invariant under addition and multiplied by |b| under scaling.

Summary (for Y = bX + a):

μ_Y = b·μ_X + a
MD_Y = |b|·MD_X
Var(Y) = b^2·Var(X)
SD_Y = |b|·SD_X
Range_Y = |b|·Range_X

Coefficient of variation (CV = SD/mean × 100): If b > 0 and a = 0, CV is unchanged by scaling. If a ≠ 0 (change of origin) CV generally changes. If mean_X = 0 then CV is undefined.

Intuition: Adding a constant shifts the whole distribution left/right without changing the spread. Multiplying by a constant stretches or shrinks distances from the origin (and from the mean), so all measures of spread scale accordingly; variance squares the scale factor because it is based on squared deviations.

What it means
Combined (or pooled) distributions arise when we join two or more frequency distributions (or samples) into a single distribution to compute overall measures of central tendency and dispersion. In grouped data we first represent observations by class intervals with frequencies and use class mid‑points as representative values.

Why we do it
To get an overall mean, variance or standard deviation for the whole population or combined samples — for example, when we want the average income of two towns taken together or the overall variability in test scores of two sections.

Key idea
Do not simply average the subgroup means: the subgroup sizes (frequencies) must be used as weights. For variance/SD, combine the sums of squared deviations (or use sums of squares about zero) before subtracting the square of the overall mean.

Procedure for grouped distributions

• Ensure classes are comparable (if different class boundaries, make a common classification).
• Find class mid‑points x_i for every class.
• Compute f_i (given), f_i x_i and f_i x_i² for every class.
• For the pooled data, add frequencies: N = Σf_i; sum of first moments: Σf_i x_i; sum of second moments: Σf_i x_i².
• Overall mean μ = (Σf_i x_i) / N.
• Overall variance σ² = (Σf_i x_i²)/N − μ². Standard deviation σ = sqrt(σ²).

Alternative formula using subgroup means and variances
When groups (or samples) are summarized by their sizes n_k, means μ_k and variances σ_k², the pooled (population) variance for K groups is
    σ_pooled² = [Σ_{k=1..K} n_k (σ_k² + μ_k²)] / N − μ_pooled²,
where μ_pooled = (Σ n_k μ_k) / N and N = Σ n_k. This works because n_k(σ_k² + μ_k²) = Σ (x²) over group k.

Note on pooled sample variance (used in inferential statistics)
If combining two independent samples and assuming equal population variance, the pooled sample variance (for use in t‑tests) is
    s_p² = [(n1−1)s1² + (n2−1)s2²] / (n1 + n2 − 2).
This is different from the population pooled variance formula above and should be used only for sample inference under the equal‑variance assumption.

Practical cautions

• Make class widths and boundaries consistent before pooling grouped data.
• Pooling can mask subgroup differences (e.g., two groups with same pooled mean may have different spreads).
• Use the sample pooled variance formula only when the equal variance assumption is justified.

Overview: Measures of dispersion describe how spread out values in a data set are around a central value. The main measures are Range, Quartile Deviation (QD or IQR/2), Mean Deviation (MD) and Standard Deviation (SD) (and Variance). A comparative study highlights their properties, strengths and limitations and gives guidance on which to use in different situations.

Comparative points:

• Definition basis: Range and QD use only extreme values or quartiles; MD uses absolute deviations from a central value (mean or median); SD uses squared deviations from the mean (all observations).
• Sensitivity to extremes: Range is most affected by outliers; QD (inter-quartile range) resists outliers; MD is moderately sensitive; SD is sensitive to outliers because of squaring.
• Mathematical properties: SD (and variance) is algebraically tractable and used in advanced analysis (e.g., inference, regression). MD has no simple algebraic simplification. QD and range are simple but less useful for further mathematical work.
• Units: Range, QD, MD and SD have same units as data. Variance has squared units.
• Ease of calculation: Range and QD are easiest; MD is easier conceptually than SD; SD is more calculation-intensive but standard tables/formulas ease computation for grouped data.
• Use with grouped data: SD and MD have well-established grouped-data formulas; QD can be found from ogives; range uses class extremes (but is imprecise if classes are wide).

When to select which measure:

• Use range for a quick, rough idea of spread (small datasets, initial checks). Not suitable when outliers are present.
• Use quartile deviation / IQR when you want a robust measure that ignores extreme values (income distribution, skewed data).
• Use mean deviation when you prefer a measure based on average absolute deviations that is easier to interpret than variance (simple comparisons, teaching).
• Use standard deviation when you need a mathematically convenient measure for further analysis (probability, normal approximation, hypothesis testing, quality control).
• For skewed distributions prefer QD/IQR or median-based MD; for symmetrical distributions SD is informative. For comparing different units or scales, use coefficient of variation (CV = SD/mean) — not a measure of dispersion alone but for relative comparison.

Practical selection rule: If robustness to outliers is important → use QD/IQR. If analytical work or comparisons based on variance are required → use SD. For quick checks or small samples → range or MD may suffice.

Overview

Measures of dispersion describe how spread out values in a data set are. Common measures: range, mean deviation (MD), quartile deviation (QD), variance and standard deviation (SD). Problem-solving means choosing the right measure, converting data appropriately (ungrouped vs grouped), and applying algebraic/shortcut methods to simplify calculation.

General problem-solving steps

1. Identify data type: ungrouped (raw list) or grouped (class intervals + frequencies).
2. Choose a measure (e.g., range for simple spread, QD for robustness, SD for mathematical work and comparison).
3. Compute a measure of central tendency needed (mean, median, quartiles).
4. For grouped data, replace each class by its class-mark (mid-point) and use frequencies.
5. Use shortcut/assumed-mean methods to reduce arithmetic where possible.
6. Check units and interpret results (e.g., compare CVs when means differ).

Techniques and shortcuts

• Shortcut (computational) formula for variance: σ² = (Σx²)/N − μ² (or for grouped: σ² = (Σf x²)/N − μ²). This avoids repeated subtraction of the mean.
• Assumed-mean method (grouped data): choose A near the mean, compute u = (x − A)/h (h = class width), then μ = A + h(Σf u)/N and σ = h * sqrt[(Σf u²)/N − (Σf u/N)²].
• For mean deviation use absolute deviations: MD = Σ|x − center|/N (or Σf|x − center|/N for grouped). Use mean or median as center—median often gives smaller MD for skewed data.
• Quartile deviation (QD) uses Q1 and Q3: QD = (Q3 − Q1)/2, robust against outliers.
• Coefficient of variation (CV) = (SD / mean) × 100% lets you compare relative dispersion for different units or scales.

Common pitfalls

• Using class limits instead of class-marks for grouped calculations; always use mid-points for x in fx and fx² unless instructed otherwise.
• For open-ended classes, choose a reasonable assumed boundary or use special techniques (CBSE problems generally avoid this complexity).
• Confusing population and sample formulas: for population divide by N; for sample variance often divide by (n−1). Class 11 problems usually use population formulas unless sample inference is introduced.

Context: In Class 11 Economics (Measures of Dispersion) we study numerical measures that describe how data are spread around a central value (mean/median/mode). Common measures: Range, Interquartile Range (IQR), Mean Deviation (MD), Variance and Standard Deviation (SD), and Coefficient of Variation (CV).

Why dispersion matters in economics: Measures of dispersion quantify uncertainty, inequality and risk — essential for comparing economic series (incomes, prices, returns, production) and for policy or business decisions.

Merits (general):

• Complement central tendency: show how representative mean/median is.
• Allow comparison of variability across different datasets (especially CV).
• Useful for risk assessment and decision making (e.g., investment risk, price volatility).
• Some measures (IQR, MD about median) are robust to extreme values.
• Standard deviation and variance have strong mathematical properties used in further statistical analysis (regression, hypothesis testing).

Limitations (general):

• Some measures are highly sensitive to outliers (Range, variance, SD).
• Variance has squared units (difficult to interpret directly); SD corrects units but still may be less intuitive than range or IQR.
• CV cannot be used if mean is zero or near zero and is meaningless for data measured on an interval scale without an absolute zero.
• All numeric dispersion measures require quantitative data; not applicable to purely qualitative data.
• Computation effort: variance/SD require squaring and summing deviations (grouped data needs midpoints and frequency handling).

Merits & limitations of specific measures:

• Range: Merits — very easy to compute and gives a quick idea of spread. Limitations — uses only two observations (min & max), extremely sensitive to outliers.
• IQR (Q3 − Q1): Merits — describes middle 50% and is robust to outliers. Limitations — ignores tails and extreme observations.
• Mean Deviation (MD): Merits — uses all observations and is easier to interpret than variance; MD about median is robust. Limitations — not algebraically convenient (absolute values), less used in advanced analysis.
• Variance & Standard Deviation (σ², σ): Merits — use all data, mathematically tractable, fundamental in probability and inferential statistics; SD is in same units as data. Limitations — variance in squared units; both sensitive to outliers.
• Coefficient of Variation (CV): Merits — allows comparison of relative variability between series with different units or means. Limitations — undefined/unstable when mean ≈ 0; not meaningful for ordinal data.

Practical applications in economics (summary):

• Income inequality: SD or CV of incomes to measure dispersion across households or regions (complements Gini coefficient).
• Risk & return in finance: Variance/SD of asset returns measures risk; CV (SD/mean) helps compare risk per unit of return across assets or funds.
• Price volatility: SD of daily/weekly prices to quantify stability of commodities or exchange rates; informs hedging and inventory decisions.
• Production and yields: CV or SD of crop yields across seasons/regions to plan storage, insurance and subsidy policies.
• Quality control and manufacturing: SD monitors consistency of product characteristics; range/IQR detect process changes or defects.
• Policy evaluation: Comparing variability of key indicators (inflation, unemployment) before/after policy interventions.
• Education assessment: SD of student marks shows dispersion in performance (helpful in targeting remedial action).

Interpretation tips:

• Always use a measure consistent with purpose: IQR for robustness, SD/variance for statistical modeling, CV to compare relative dispersion.
• Plot data (boxplot, histogram, time series) before relying on a single numerical measure — visual checks reveal skewness and outliers.

Overall: Measures of dispersion are indispensable in economics for measuring inequality, volatility and risk. Choice of measure depends on sensitivity to extremes, interpretability, and the need for algebraic properties in further analysis.