One-Variable Statistics: Summary Measures, Spread, and Distributions

Classification and Representation of Data

  • Qualitative vs. Quantitative Data:

    • Qualitative (Categorical): Descriptive observations. Subdivided into Nominal (unordered categories, e.g., academic programs) and Ordinal (categories with an inherent ranking or order, e.g., skill levels).

    • Quantitative (Numerical): Numerical counts or measurements. Subdivided into Discrete (isolated, countable values) and Continuous (measurable real numbers over an interval).

  • Data Sources and Forms:

    • Primary Data: Collected directly by the researcher.

    • Secondary Data: Collected by an external source.

    • Experimental vs. Observational: Data derived from manipulated statistical models versus observed natural occurrences.

    • Microdata vs. Aggregate Data: Raw individual data points versus summarized, processed figures.

  • Tabular Summary Metrics:

    • Frequency (ff): Number of times a value occurs.

    • Relative Frequency (rfrf): Proportion of total occurrences, given by rf=fn×100%rf = \frac{f}{n} \times 100\%.

    • Cumulative Frequency (cfcf): Running total of frequencies up to and including a specific interval.

    • Cumulative Relative Frequency (crfcrf): Cumulative frequency as a percentage of the total dataset, given by crf=cfn×100%crf = \frac{cf}{n} \times 100\%.

  • Graphical Displays:

    • Bar Graphs: Used exclusively for discrete data categories.

    • Histograms: Used for continuous data grouped into class intervals.

    • Frequency Polygons: Applicable to both discrete and continuous data; facilitates distribution comparison across multiple sets.

    • Time Series Plots: Illustrates changes in a single variable across successive time periods.

Measures of Central Tendency

  • Mean (xˉ\bar{x}):

    • The arithmetic average, computed for grouped or frequency data as:     xˉ=∑fx∑f\bar{x} = \frac{\sum fx}{\sum f}

    • Highly sensitive to extreme values and outliers.

  • Weighted Mean (xˉ\bar{x}):

    • Applied when data values hold unequal significance or differing assigned weights (ww):     xˉ=∑wx∑w\bar{x} = \frac{\sum wx}{\sum w}

  • Median:

    • The middle value in a sorted dataset, located at the rank position:     Position=n+12\text{Position} = \frac{n + 1}{2}

    • Resistant to outliers; preferred indicator of central tendency in skewed data.

  • Mode:

    • The data value (xx) or class interval associated with the highest frequency (ff). Resistant to outliers.

Measures of Spread and Position

Hypothetical histograms showing wide and narrow spreads about a central value
  • Range:

    • The total distance between the extremes of a dataset:     Range=Highest value−Lowest value\text{Range} = \text{Highest value} - \text{Lowest value}

  • Quartiles:

    • Divide an ordered dataset into four equal segments:

    • First Quartile (Q1Q_1): Located at the position n+14\frac{n + 1}{4}.

    • Second Quartile (Q2Q_2 / Median): Located at the position n+12\frac{n + 1}{2}.

    • Third Quartile (Q3Q_3): Located at the position 3(n+1)4\frac{3(n + 1)}{4}.

  • Interquartile Range (IQRIQR):

    • Measures the spread of the middle 50%50\% of observations, removing outlier distortion:     IQR=Q3−Q1IQR = Q_3 - Q_1

    • Semi-Interquartile Range: Half of the IQRIQR, defined as IQR2\frac{IQR}{2}.

  • Box-and-Whisker Plot (Five-Number Summary):

Box-and-whisker plot diagram displaying the five-number summary and range
  • Visually displays the minimum, Q1Q_1, median, Q3Q_3, and maximum.

  • Outliers are represented as discrete points beyond the whiskers.

    • Percentiles:

  • Positional indicators defining the percentage of observations that a particular score exceeds (distinct from percentage scores).

Variance, Standard Deviation, and Z-Scores

  • Deviations:

    • The distance of an observation from the mean (x−xˉx - \bar{x} or x−μx - \mu). The algebraic sum of all deviations across a dataset equals 00.

  • Variance:

    • Mean of the squared deviations from the central mean.

    • Population Variance: σ2=∑f(x−μ)2N\sigma^2 = \frac{\sum f(x - \mu)^2}{N}

    • Sample Variance: s2=∑f(x−xˉ)2n−1s^2 = \frac{\sum f(x - \bar{x})^2}{n - 1}

    • (Note: Division by n−1n - 1 corrects for sample underestimation of population variability).

  • Standard Deviation:

    • The square root of the variance, matching the units of the original dataset:     σ=∑f(x−μ)2N\sigma = \sqrt{\frac{\sum f(x - \mu)^2}{N}}     s=∑f(x−xˉ)2n−1s = \sqrt{\frac{\sum f(x - \bar{x})^2}{n - 1}}

  • Z-Scores:

    • Indicates how many standard deviations a data point lies above or below the mean:     z=x−μσz = \frac{x - \mu}{\sigma}     z=x−xˉsz = \frac{x - \bar{x}}{s}

    • Enables objective standardized comparisons between sets containing different averages and variations.

Distribution Shapes and Skewness

Normal and skewed unimodal distribution curves
  • Uniform (Rectangular) Distribution: All values possess equal frequencies across the entire domain.

  • U-Shaped Distribution: Frequencies peak at the lower and upper bounds, with low frequency in the center.

  • Bimodal / Multimodal Distribution: Features two or more prominent local frequency peaks.

  • Normal (Symmetric) Distribution: Bell-shaped curve centered at the mean, where:   Mean=Median=Mode\text{Mean} = \text{Median} = \text{Mode}

  • Positively Skewed (Skewed Right):

    • The tail extends toward the right (positive direction); observations cluster toward the lower end.

    • Order of central metrics: Mode<Median<Mean\text{Mode} < \text{Median} < \text{Mean}.

    • Box plot features a longer right whisker, with the median positioned left of center.

  • Negatively Skewed (Skewed Left):

    • The tail extends toward the left (negative direction); observations cluster toward the higher end.

    • Order of central metrics: Mean<Median<Mode\text{Mean} < \text{Median} < \text{Mode}.

    • Box plot features a longer left whisker, with the median positioned right of center.

The Normal Distribution

Normal distribution empirical rule percentage breakdown
  • Empirical Rule (68–95–99.7%68\text{--}95\text{--}99.7\%):

    • Approximately 68%68\% of observations lie within ±1σ\pm 1\sigma (±1s\pm 1s) of the mean (34%34\% on either side).

    • Approximately 95%95\% of observations lie within ±2σ\pm 2\sigma (±2s\pm 2s) of the mean (13.5%13.5\% between 1σ1\sigma and 2σ2\sigma on each side).

    • Approximately 99–100%99\text{--}100\% (99.7%99.7\%) of observations lie within ±3σ\pm 3\sigma (±3s\pm 3s) of the mean.

    • Tail regions beyond ±2σ\pm 2\sigma encompass 2.5%2.5\% of data in each respective tail.