Descriptive Statistics, Central Tendency, Dispersion & Normal Curve

In psychological research and psychometrics, quantitative analysis transforms raw behavioral observations into meaningful, interpretable scientific conclusions. Descriptive statistics summarize, organize, and describe the characteristics of a specific sample, establishing the empirical foundation for subsequent inferential testing.

For CUET PG Psychology (HUQP20), questions in this domain frequently test:

S.S. Stevens' Measurement Scales (Nominal, Ordinal, Interval, Ratio) and their permissible mathematical operations and statistical tests.
Frequency Distributions and Graphical Techniques (Histograms, Frequency Polygons, Cumulative Ogives, and Bar Charts).
Measures of Central Tendency (Mean, Median, Mode: mathematical properties, balance points, and empirical relationships).
Measures of Dispersion / Variability (Range, Quartile Deviation, Mean Deviation, Variance, and Standard Deviation with Bessel's correction).
The Normal Probability Curve (NPC / Gaussian Distribution): Mathematical properties, points of inflection, standard score transformations ( $z, T$ ), and fractional area calculations.
Divergence from Normality: Mathematical definitions, causes, and diagnostic metrics of Skewness and Kurtosis.

1. Epistemological Foundations: Scales of Measurement

In 1946, S.S. Stevens published a landmark taxonomy in Science, defining measurement as the "assignment of numerals to objects or events according to rules." Stevens identified four distinct levels of measurement based on four mathematical properties: Identity (Categorization), Order (Magnitude), Equal Intervals, and Absolute (True) Zero.

Scale Level	Defining Mathematical Properties	True Absolute Zero	Permissible Central Tendency & Dispersion	Valid Statistical Tests	Psychological Examples
Nominal	Identity only ( $A = B$ or $A \neq B$ ). Numbers serve strictly as qualitative labels.	No	Mode, Frequency counts, Percentages	Chi-Square ( $\chi^2$ ), Contingency Coefficient	Diagnostic codes (DSM-5), biological sex, experimental group codes (0/1).
Ordinal	Identity + Magnitude ( $A > B > C$ ). Distance between ranks is unequal and unknown.	No	Median, Interquartile Range ( $IQR$ ), Quartile Deviation ( $QD$ )	Spearman's $\rho$ , Mann-Whitney $U$ , Wilcoxon signed-rank	Likert attitude scales, socioeconomic status (SES), percentile ranks ( $PR$ ).
Interval	Identity + Magnitude + Equal Intervals. Quantitative distance is meaningful.	No (Arbitrary zero)	Mean, Standard Deviation ( $\sigma, s$ ), Variance ( $s^2$ )	Pearson correlation ( $r$ ), Student's $t$ -test, ANOVA ( $F$ -test)	Standardized intelligence tests (WAIS IQ), temperature ( $^\circ\text{C}, ^\circ\text{F}$ ), neuroticism scores.
Ratio	Identity + Magnitude + Equal Intervals + Absolute Zero (complete absence).	Yes	Geometric Mean, Harmonic Mean, Coefficient of Variation	All mathematical & parametric operations ( $+, -, \times, \div$ )	Reaction time (ms), error tallies, distance in maze, galvanic skin response.

1. Nominal Scale

Definition: Numbers or symbols serve strictly as qualitative labels or tags to classify objects into mutually exclusive and exhaustive categories.
Mathematical Property: Identity only ( $A = B$ or $A \neq B$ ). Numbers have no quantitative value or rank.
Psychological Examples: Psychiatric diagnostic codes (e.g., DSM-5 code 295.90 for Schizophrenia), biological sex (1 = Male, 2 = Female), experimental group assignment (0 = Control, 1 = Treatment).
Permissible Operations: Counting frequencies, percentages, modal category, and non-parametric contingency tests ( $\chi^2$ ). Arithmetic operations ( $\bar{X}, SD$ ) are strictly meaningless.

2. Ordinal Scale

Definition: Quantifies individuals along an ordered continuum; scores reflect relative ranking or magnitude.
Mathematical Property: Identity and Order ( $A > B > C$ ). However, the distance between successive ranks is unknown and unequal (the difference between rank 1 and 2 is not necessarily equal to that between rank 2 and 3).
Psychological Examples: Likert response scales (Strongly Disagree to Strongly Agree), socioeconomic status (Lower, Middle, Upper), class rank, percentile ranks ( $PR$ ).
Permissible Operations: Median, interquartile range ( $IQR$ ), rank-order correlations (Spearman's $\rho$ , Kendall's $\tau$ ), non-parametric tests (Mann-Whitney $U$ , Wilcoxon signed-rank).

3. Interval Scale

Definition: Possesses quantitative order and equal numerical intervals between adjacent scale units, but lacks an absolute (true) zero point.
The Arbitrary Zero: Zero does not signify the complete absence of the measured attribute; it is an arbitrary convention. For example, $0^\circ\text{C}$ does not represent the absence of molecular heat. Consequently, ratio statements are invalid (e.g., $80^\circ\text{F}$ is not "twice as hot" as $40^\circ\text{F}$ ).
Psychological Examples: Standardized intelligence scores (Wechsler IQ, Stanford-Binet; an individual with an IQ of 140 is not "twice as smart" as one with an IQ of 70, nor does an IQ of 0 imply zero intelligence), temperature (Celsius/Fahrenheit), personality inventory scale scores.
Permissible Operations: Addition and subtraction; Mean, Standard Deviation, Pearson Product-Moment Correlation ( $r$ ), Student's $t$ -test, Analysis of Variance ( $F$ -test).

4. Ratio Scale

Definition: Possesses all the properties of an interval scale alongside a true, non-arbitrary, absolute zero point representing the complete absence of the construct.
Ratio Validity: Because zero is absolute, ratios of numbers reflect authentic physical ratios ( $10\text{ seconds}$ is exactly twice as long as $5\text{ seconds}$ ).
Psychological Examples: Reaction time (RT) in milliseconds, error counts on an operant conditioning task, distance traveled in an open-field maze, galvanic skin conductance amplitude.
Permissible Operations: All mathematical operations ( $+, -, \times, \div$ ), geometric mean, harmonic mean, and relative dispersion measures (Coefficient of Variation).

2. Organization and Graphical Representation of Data

When raw psychological scores are collected, they are structured into grouped frequency distributions by establishing class intervals.

Concept	Technical Definition	Mathematical Operationalization	Empirical Psychometric Rule
Class Interval ( $i$ )	Width or score span grouped together.	$i = \text{Upper Real Limit} - \text{Lower Real Limit}$	Typically $10\text{--}20$ intervals are chosen to balance detail with data reduction.
Stated (Apparent) Limits	Score boundaries as conventionally written.	E.g., $50\text{--}54$ , $55\text{--}59$	Discrete integer boundaries reported on data collection sheets.
Real (Exact) Limits	Continuous theoretical score boundaries.	Lower: $\text{Stated Lower} - 0.5$ ; Upper: $\text{Stated Upper} + 0.5$	For interval $50\text{--}54$ , exact limits are $49.5\text{--}54.5$ .
Class Midpoint ( $X$ )	Central balance point representing the interval.	$X = \frac{\text{Lower Limit} + \text{Upper Limit}}{2}$	Used as the single quantitative coordinate in frequency polygons.

Primary Graphical Modalities

Graphical Modality	Coordinate Construction ( $X$ -axis / $Y$ -axis)	Geometric Morphology	Core Psychometric Function
Histogram	$X$ : Exact Real Limits; $Y$ : Frequency ( $f$ )	Contiguous vertical columns; no gaps between bars; total area equals sample size ( $N$ ).	Visualizes shape of continuous frequency distributions and modality.
Frequency Polygon	$X$ : Class Midpoints ( $X$ ); $Y$ : Frequency ( $f$ )	Closed line graph; anchored to baseline at hypothetical midpoints one interval below lowest and above highest.	Ideal for comparing multiple distributions simultaneously on identical axes.
Cumulative Ogive	$X$ : Upper Exact Limits; $Y$ : Cumulative Frequency ( $cf$ ) or Cumulative % ( $c\%$ )	Smooth S-shaped (sigmoidal) ascending curve; reaches $100\%$ asymptotically.	Rapid visual estimation of Median ( $P_{50}$ ), quartiles ( $Q_1, Q_3$ ), and percentile ranks ( $PR$ ).

Histogram: A set of contiguous vertical columns constructed over the horizontal axis ( $X$ -axis), where the width of each column corresponds to the exact real limits of the class interval and the height represents frequency ( $f$ ). No gaps exist between bars, reflecting the continuous nature of the variable.
Frequency Polygon: A line graph formed by plotting frequencies directly above the midpoint ( $X$ ) of each class interval and connecting adjacent points with straight lines.
- Anchor Rule: To close the figure and ensure the total area beneath the polygon equals the total frequency ( $N$ ), the line is extended to touch the horizontal baseline at the midpoints of the hypothetical intervals immediately preceding the lowest and succeeding the highest interval.
Cumulative Frequency Curve (Ogive): Plotted by calculating cumulative frequencies ( $cf$ $c f$ ) or cumulative percentages ( $c\%$ $c %$ ) and plotting them against the upper real limit of each class interval.
- Shape: Yields a characteristic smooth S-shaped (sigmoidal) curve.
- Psychometric Utility: Critical for determining median ( $50\text{th percentile}$ ), quartiles ( $Q_1, Q_3$ ), and percentile ranks ( $PR$ ).

3. Measures of Central Tendency

Measures of central tendency represent single representative numerical values around which the scores of a distribution cluster.

Measure	Fundamental Definition	Defining Mathematical Hallmark	Sensitivity to Distribution Shape
Arithmetic Mean ( $\bar{X}$ )	Arithmetic balance point; sum of values divided by count ( $N$ ).	Center of gravity: $\sum (X - \bar{X}) = 0$ ; Least squares: $\sum (X - \bar{X})^2 = \text{Minimum}$ .	Pulled heavily toward the long tail in skewed distributions; sensitive to outliers.
Median ( $Mdn$ )	Positional midpoint; $50\text{th percentile}$ ( $P_{50}$ ) dividing data into two equal halves.	Divides ranked observations into equal frequencies ( $50\%$ above, $50\%$ below).	Highly robust; unaffected by extreme outliers or open-ended intervals.
Mode ( $Mo$ )	Most frequently occurring score or category.	Peak ordinate of distribution curve; Pearson formula: $\text{Mode} = 3(Mdn) - 2(\bar{X})$ .	Crudest index; can be unimodal, bimodal, or multimodal; ignores actual score magnitudes.

1. The Arithmetic Mean ( $\bar{X}$ or $\mu$ )

Formulas: $\text{Ungrouped Data: } \bar{X} = \frac{\sum X}{N} \qquad\qquad \text{Grouped Data: } \bar{X} = \frac{\sum fX}{N} = AM + \left(\frac{\sum fx'}{N}\right) \times i$ (where $AM$ is assumed mean, $x' = \frac{X - AM}{i}$ is coded step deviation).
Crucial Mathematical Properties:
1. Balance Point: The sum of algebraic deviations of all raw scores from the mean is strictly zero: $\sum (X - \bar{X}) = 0$
2. Least Squares Criterion: The sum of squared deviations of scores from the mean is smaller than from any other reference point: $\sum (X - \bar{X})^2 = \text{Minimum}$
3. Linear Sensitivity: Adding or multiplying a constant $k$ to every score changes the mean by that exact constant: $\text{New Mean} = \bar{X}_{\text{old}} \pm k \quad\text{or}\quad \bar{X}_{\text{old}} \times k$
4. Vulnerability to Outliers: The mean is pulled substantially toward the tail of a skewed distribution.

2. The Median ( $Mdn$ )

Definition: The point on the score scale below which exactly $50\%$ of the observations fall ( $P_{50}$ ).
Formulas:
- Ungrouped Data: Arrange in ascending order. If $N$ is odd, position is $\frac{N+1}{2}$ . If $N$ is even, it is the arithmetic average of the two middle values.
- Grouped Data: $Mdn = L + \left[\frac{\frac{N}{2} - F}{f_m}\right] \times i$ (where $L$ = lower exact limit of median class, $F$ = cumulative frequency below $L$ , $f_m$ = frequency of median class, $i$ = class size).
Core Characteristic: Highly resistant (robust) to extreme outliers and open-ended distributions.

3. The Mode ( $Mo$ )

Definition: The score or category that occurs with the highest frequency in a distribution.
Empirical Formula (Karl Pearson): In moderately skewed distributions, the mode can be approximated from the mean and median: $\text{Mode} = 3(\text{Median}) - 2(\text{Mean})$ $\text{Rearranged: } \text{Mean} - \text{Mode} = 3(\text{Mean} - \text{Median})$

4. Measures of Dispersion / Variability

Central tendency indicates where a distribution centers; dispersion (variability) measures the extent to which individual scores scatter, spread, or deviate from that central point. Two distributions can have identical means yet entirely different dispersions.

[!NOTE] Homogeneous vs. Heterogeneous Distribution Illustration:

Distribution A: $48, 50, 52 \implies \text{Mean} = 50, \text{Range} = 4, SD = 2.0$ (Highly homogeneous)

Distribution B: $10, 50, 90 \implies \text{Mean} = 50, \text{Range} = 80, SD = 40.0$ (Highly heterogeneous)

Measure of Dispersion	Mathematical Formula	Key Characteristics & Psychometric Role
Range ( $R$ )	$R = X_{\text{max}} - X_{\text{min}} (+ 1)$	Crudest index; considers only two extreme scores; highly sensitive to sample size fluctuations.
Quartile Deviation ( $QD$ )	$QD = \frac{Q_3 - Q_1}{2}$	Semi-interquartile range; encompasses central $50\%$ of cases; paired with Median for skewed/ordinal data.
Mean Deviation ( $MD$ )	$MD = \frac{\sum \\|X - \bar{X}\\|}{N}$	Arithmetic average of absolute linear deviations; algebraically inflexible; rarely used in inferential tests.
Sample Variance ( $s^2$ )	$s^2 = \frac{\sum (X - \bar{X})^2}{N - 1}$	Mean squared deviation; incorporates Bessel's correction ( $N-1$ ); foundational additive metric for ANOVA.
Standard Deviation ( $s, \sigma$ )	$s = \sqrt{\frac{\sum (X - \bar{X})^2}{N - 1}}$	Root-mean-square deviation; expressed in original scale units; universal unit of normal distribution and $z$ -scores.