Discrete Probability Distributions

Core Statistical Definitions and Variable Classifications

  • Variable Definition:

    • A variable is any characteristic of an individual, object, or entity that can take on different numerical values or be assigned to distinct categories.

  • Variable Classifications:

    • Categorical Variable: A variable that classifies observations into distinct, non-numerical groups or quality classes (e.g., color, gender, blood type).

    • Ordinal Variable: A specific subtype of categorical variable whose categories have an inherent, meaningful rank order or hierarchy (e.g., low, medium, high; stage of disease).

    • Quantitative Variable: A numerical variable that measures and records a countable amount or quantifiable magnitude for an observation.

    • Continuous Variable: A quantitative variable measured on a continuous, uninterrupted scale capable of taking on an infinite number of possible fractional or decimal values within a given interval (e.g., height, weight, concentration).

    • Discrete Variable: A quantitative variable with a countable set of distinct, isolated values, with no possible values between consecutive integers (e.g., counts of events, number of offspring).

  • Sample and Sample Size:

    • Statistical Sample: A defined collection of objects or individuals drawn from a broader population, upon which one or more variables are measured.

    • Sample Size: The total number of observed entities within a sample, denoted statistically by the lowercase letter nn.

    • Biological versus Statistical Terminology Distinction: In laboratory biological contexts, the word "sample" often refers to a physical biological specimen (e.g., an individual blood tube, tissue aliquot, or biopsy). A biological "blood sample" represents only a single measurement unit (n=1n = 1) and does not constitute a statistical sample of a population.

Frequency Distributions and Categorical Data Visualization

  • Frequency Distribution Fundamentals:

    • A frequency distribution is a tabular, graphical, or mathematical depiction that displays the number of occurrences (frequencies) of each individual value or category in a dataset.

    • Distributions can be expressed in data tables, plotted on specialized charts, or modeled using continuous mathematical probability functions.

  • Tabular Frequency Distribution Case Study (62 oz Jar of Peanut M&Ms):

    • Total observation count: n=722n = 722

    • Frequency counts across observed color classes:

    • Red: 8888

    • Blue: 171171

    • Orange: 161161

    • Yellow: 105105

    • Green: 113113

    • Brown: 8484

    • Total Count: 722722

  • Bar Graphs (Column Graphs):

    • Well suited for displaying counts and frequencies of discrete quantitative variables and nominal categorical categories.

    • Displays discrete categories along the horizontal axis (XX-axis) separated by distinct gaps, emphasizing that categories are separate and non-continuous.

    • Design and Publication Considerations: When creating bar charts using statistical software (such as Microsoft Excel), consider the reproduction format; charts designed with distinct colors must remain interpretable if converted to monochrome or grayscale print.

  • Pie Charts and Proportional Composition:

    • Used when the primary objective is to represent proportional shares or percentages of a whole rather than raw frequency counts.

    • Each slice of the circle corresponds to the percentage of the total sample size accounted for by that category.

    • Proportional distribution for the 62 oz62\text{ oz} Peanut M&Ms dataset:

    • Blue: 24%24\%

    • Orange: 22%22\%

    • Green: 16%16\%

    • Yellow: 14%14\%

    • Red: 12%12\%

    • Brown: 12%12\%

    • Caption Requirements: Scientific standards require every graphic, chart, or visual representation to feature an explicit caption explaining the subject matter, sample context, and experimental parameters.

Graphical representation of the frequency of colors in a 62 oz. container of Peanut M&Ms

Graphical Displays for Small and Continuous Datasets

  • Dotplots:

    • Used for displaying quantitative data distributions when the sample size nn is small.

    • Each dot plotted above the horizontal scale represents a single individual observation.

    • Can be constructed in statistical packages such as Minitab to identify data density, clusters, and spread across a numerical scale.

    • Example Case: Infant mortality rate per 10,00010{,}000 population across 1111 third-world nations, plotting values at approximately 1010, 1414, 1818, 2525, 2727, 3131, 3636, 3737, 3939, 4949, and 6262.

Dotplot of mortality rate in 11 countries
  • Stem-and-Leaf Diagrams:

    • Useful for organizing and visualizing small datasets without discarding the raw numerical values of individual observations.

    • Structure:

    • Stem: Represents the leading digit or digits (e.g., tens place).

    • Leaf: Represents the trailing digit (e.g., units place).

    • The horizontal length of each leaf line directly reflects the concentration of observations at that level of the stem.

    • Dataset: Radish Growth in Millimeters (n=20n = 20):

    • Raw values: 15,20,11,30,33,16,29,35,8,10,17,37,15,25,23,18,26,21,19,2615, 20, 11, 30, 33, 16, 29, 35, 8, 10, 17, 37, 15, 25, 23, 18, 26, 21, 19, 26

    • Formatted Stem-and-Leaf Plot:

      • 0∣80 \mid 8

      • 1∣0 11 \mid 0\, 1

      • 1∣5 5 6 7 8 91 \mid 5\, 5\, 6\, 7\, 8\, 9

      • 2∣0 1 32 \mid 0\, 1\, 3

      • 2∣5 6 6 92 \mid 5\, 6\, 6\, 9

      • 3∣0 33 \mid 0\, 3

      • 3∣5 73 \mid 5\, 7

  • Histograms:

    • Displays continuous quantitative variables that possess an inherent order, scale, and continuous range.

    • Constructed by partitioning the continuous range into adjacent, non-overlapping intervals (bins) and plotting the count or frequency of observations falling into each interval.

    • The tick mark along the horizontal axis indicates the exact midpoint of each interval bin.

    • Unlike bar graphs for discrete categories, the vertical columns of a histogram touch each other without spaces, illustrating the uninterrupted continuity of the underlying variable.

Histogram of continuous height data

Distribution Characteristics: Normality, Skewness, and Kurtosis

  • Evaluating Continuous Data for Normality:

    • Histograms are frequently inspected to determine whether sample data conform to a theoretical normal distribution (a symmetric, bell-shaped Gaussian curve).

    • Normality is an underlying assumption required for many standard parametric statistical tests.

Histogram displaying a bell-shaped normal distribution curve
  • Classification of Distribution Shapes:

    • Symmetric, Bell-Shaped: Balanced distribution where the left and right halves are mirror images, peaking in the center.

    • Symmetric, Non-Bell-Shaped: Balanced mirror distribution that diverges from the standard bell profile (e.g., platkurtic flat distributions or sharp spikes with extended flat shoulders).

    • Skewed to the Right (Positive Skew): Asymmetric distribution with the bulk of observations concentrated at lower values and a long, tapering tail extending rightward toward positive infinity.

    • Skewed to the Left (Negative Skew): Asymmetric distribution where the bulk of observations reside at higher values and a long, tapering tail extends leftward toward lower values.

    • Exponential Distribution: Distribution characterized by an immediate peak at zero or the lowest boundary, decaying monotonically toward the right.

    • Bimodal Distribution: Distribution displaying two distinct local peaks (modes), typically indicating that the sample combines two distinct subpopulations.

Types of distribution shapes including symmetric, skewed, exponential, and bimodal
  • Skewness and Kurtosis Metrics:

    • Skewness: Quantifies the directional asymmetry of the distribution. Positive values indicate a rightward tail, negative values indicate a leftward tail, and zero denotes symmetry.

    • Kurtosis: Quantifies the tailedness and peakedness of the distribution relative to a standard normal curve. High kurtosis reflects heavy tails and a sharp central peak; low kurtosis indicates light tails and a flattened profile.

  • Sucrose Concentration Dataset (n=900n = 900 observations):

    • Total valid observations (NN): 900900

    • Missing values (N∗N^*): 00

    • Mean: 14.93614.936

    • Standard Error of the Mean (SE Mean): 0.1810.181

    • Standard Deviation (StDev): 5.4415.441

    • Minimum: 0.4040.404

    • First Quartile (Q1Q_1): 11.32311.323

    • Median: 14.79314.793

    • Third Quartile (Q3Q_3): 18.36418.364

    • Maximum: 35.37635.376

    • Skewness: 0.300.30

    • Kurtosis: 0.420.42

Histogram of sucrose concentration distribution

Bivariate Data and Regression Analysis

  • Evaluating Relationships Between Paired Variables:

    • Bivariate analysis investigates how two separate quantitative variables, measured simultaneously on the same individual subjects, relate to one another.

    • Paired Dataset Example: Golden retriever litters comparing litter size (independent variable) against mean puppy weight in pounds (dependent variable):

    • Litter size 11: Mean weight 2.4 lbs2.4\,\text{lbs}

    • Litter size 22: Mean weight 2.2 lbs2.2\,\text{lbs}

    • Litter size 33: Mean weight 2.1 lbs2.1\,\text{lbs}

    • Litter size 44: Mean weight 1.9 lbs1.9\,\text{lbs}

    • Litter size 55: Mean weight 1.9 lbs1.9\,\text{lbs}

    • Litter size 66: Mean weight 1.6 lbs1.6\,\text{lbs}

    • Litter size 77: Mean weight 1.5 lbs1.5\,\text{lbs}

    • Litter size 88: Mean weight 1.2 lbs1.2\,\text{lbs}

  • Scatter Plot Principles:

    • The independent variable (explanatory variable) is always plotted on the horizontal axis (XX-axis).

    • The dependent variable (response variable) is always plotted on the vertical axis (YY-axis).

    • Point Connection Rule: Points in a scatter plot must not be connected with line segments. Connecting the dots falsely implies continuity between discrete litters where no intermediate data points exist.

  • Linear Regression and Line of Best Fit:

    • A linear regression line (line of best fit) can be fitted mathematically across the scatter plot to model and summarize the functional relationship between the variables.

    • Visual displays require standardized captions for scientific reporting (e.g., "Fig. 2. Line graph illustrating the relation between litter size and mean weight (in pounds) in litters of golden retrievers.").

Scatter plot with linear regression line of best fit for litter size and mean weight

Parameters Versus Statistics

  • Fundamental Distinctions:

    • Population Parameter:

    • A descriptive numerical measure that characterizes an entire population.

    • Parameters are fixed, unchanging constants for a given population at a specific point in time.

    • Symbolized using Greek letters, such as μ\mu for population mean and σ\sigma for population standard deviation.

    • Sample Statistic:

    • A descriptive numerical measure computed from an empirical sample drawn from a population.

    • Statistics are random variables that fluctuate across different independent samples drawn from the same underlying population.

    • Designed to serve as point estimators for unknown population parameters.

    • Symbolized using Latin/English letters typically chosen from the end of the alphabet (e.g., xx, yy, Yˉ\bar{Y}, ss, s2s^2).

  • Primary Descriptors of Quantitative Data:

    • Central Tendency: Identifies the single central value or location around which empirical data points cluster (the "middle" of the distribution).

    • Dispersion (Variability): Quantifies the magnitude of spread, scatter, or degree of difference among individual observations within the dataset.

Measures of Central Tendency

  • Definitions of Central Measures:

    • Median: The exact physical middle value when observations are sorted in ascending order. If sample size nn is odd, it is the center observation; if nn is even, it is the arithmetic average of the two central observations.

    • Mid-Range: The algebraic average of the two extreme boundary observations in the sample:

Mid-Range=Minimum+Maximum2\text{Mid-Range} = \frac{\text{Minimum} + \text{Maximum}}{2}

  • Mode: The specific observed value that appears with the highest frequency in the dataset. A distribution can be unimodal, bimodal, multimodal, or have no mode if all values are unique.

  • Mean: The arithmetic average that mathematically minimizes the sum of squared deviations around it. The mean is the only measure of central tendency whose calculation uses every single observation in the data set.

    • Mathematical Formulations for the Mean:

  • Population Mean (μ\mu): Calculated over a population of total size NN:

μ=∑i=1NxiN\mu = \frac{\sum_{i=1}^N x_i}{N}

  • Sample Mean (Yˉ\bar{Y} or Xˉ\bar{X}): Calculated over an empirical sample of size nn:

Yˉ=∑i=1nXin\bar{Y} = \frac{\sum_{i=1}^n X_i}{n}

  • Worked Comparison Samples:

    • Sample 1: 4,5,5,6,84, 5, 5, 6, 8 (n=5n = 5

    • Sample 2: 4,6,7,7,8,94, 6, 7, 7, 8, 9 (n=6n = 6

  • Software Implementation (Excel and Minitab):

    • Microsoft Excel:

    • Basic formulas: =AVERAGE(range), =MEDIAN(range), =MODE(range).

    • Data Analysis ToolPak: Using the "Descriptive Statistics" tool provides an automated summary table.

    • Minitab Navigation Path: Enter data into the spreadsheet columns, select the Stat tab, click Basic Statistics, and select Display Descriptive Statistics.

Excel worksheet showing summary measures of central tendency for Sample 1 and Sample 2Minitab output displaying descriptive statistics for Sample 1 and Sample 2
  • Full Summary Output for Samples 1 and 2:

    • Sample 1 Summary Statistics:

    • Mean: 5.65.6

    • Standard Error: 0.6782330.678233

    • Median: 55

    • Mode: 55

    • Standard Deviation: 1.5165751.516575

    • Sample Variance: 2.32.3

    • Kurtosis: 1.4555771.455577

    • Skewness: 1.118081.11808

    • Range: 44

    • Minimum: 44

    • Maximum: 88

    • Sum: 2828

    • Count (nn): 55

    • Mid-Range: 6.06.0

    • Mode Frequency (N for ModeN\text{ for Mode}): 22

    • Sample 2 Summary Statistics:

    • Mean: 6.8333336.833333

    • Standard Error: 0.7031670.703167

    • Median: 77

    • Mode: 77

    • Standard Deviation: 1.7224011.722401

    • Sample Variance: 2.9666672.966667

    • Kurtosis: 0.8142910.814291

    • Skewness: −0.67844-0.67844

    • Range: 55

    • Minimum: 44

    • Maximum: 99

    • Sum: 4141

    • Count (nn): 66

    • Mid-Range: 6.56.5

    • Mode Frequency (N for ModeN\text{ for Mode}): 22

Measures of Dispersion and Variability

  • Measures of Dispersion:

    • Dispersion describes how spread out observations are around the central tendency.

    • The four primary measures of dispersion are:

    • Range

    • Variance

    • Standard Deviation

    • Quartiles

  • Population Variance (σ2\sigma^2):

    • Conceptual Formula: Measures the average squared deviation of all population observations from the true population mean μ\mu:

σ2=∑i=1N(xi−μ)2N\sigma^2 = \frac{\sum_{i=1}^N (x_i - \mu)^2}{N}

  • Computational / Calculator Shortcut Formula:

σ2=∑i=1NX2−(∑i=1NX)2NN\sigma^2 = \frac{\sum_{i=1}^N X^2 - \frac{\left(\sum_{i=1}^N X\right)^2}{N}}{N}

  • Population Standard Deviation (σ\sigma):

    • Because variance is expressed in squared units of the original measurement, taking the square root restores the metric to the original units:

σ=σ2\sigma = \sqrt{\sigma^2}

  • Sample Variance (s2s^2):

    • Sample variance is a variable; each unique sample pulled from a population yields a different variance estimate.

    • Degrees of Freedom Correction (n−1n - 1): When calculating sample variance, deviations are measured from the estimated sample mean Xˉ\bar{X} rather than the true population parameter μ\mu. Dividing by nn results in an underestimate (biased estimate) of the true population variance. Dividing by the degrees of freedom (n−1n - 1) corrects this bias, making s2s^2 an unbiased estimator of σ2\sigma^2.

    • Conceptual Definition: The average squared deviation from the sample mean.

    • Conceptual Formula:

s2=∑i=1n(Xi−Xˉ)2n−1s^2 = \frac{\sum_{i=1}^n (X_i - \bar{X})^2}{n - 1}

  • Computational / Calculator Shortcut Formula:

s2=∑i=1nX2−(∑i=1nX)2nn−1s^2 = \frac{\sum_{i=1}^n X^2 - \frac{\left(\sum_{i=1}^n X\right)^2}{n}}{n - 1}

  • Sample Standard Deviation (ss):

    • Conceptual Definition: The square root of the average squared deviation from the sample mean.

    • Returns dispersion to the original measurement units:

s=s2s = \sqrt{s^2}

Quartiles and Boxplots

  • Quartile Definitions:

    • Quartiles divide an ordered dataset into four equal segments, each containing 25%25\% of the total observations:

    • First Quartile (Q1Q_1): The 25th percentile; separates the lowest 25%25\% of data from the upper 75%75\%.

    • Second Quartile (Q2Q_2): The 50th percentile, corresponding to the sample median.

    • Third Quartile (Q3Q_3): The 75th percentile; separates the lower 75%75\% of data from the top 25%25\%.

    • Interquartile Range (IQR): The difference between the third and first quartiles, containing the central 50%50\% of the data:

IQR=Q3−Q1\text{IQR} = Q_3 - Q_1

  • Boxplot (Box-and-Whisker Plot) Structure:

    • Provides a graphical summary of a continuous dataset based on a five-number summary:

    • Central Box: Spans from Q1Q_1 to Q3Q_3, representing the middle 50%50\% of the dataset.

    • Center Line: A horizontal cross-line within the box indicating the position of the median (Q2Q_2).

    • Whiskers: Vertical lines extending from the edges of the box outward to the minimum and maximum observed values (or defined non-outlier limits).

  • Boxplot Worked Examples:

    • Example 1 Dataset: 2,2,4,5,8,9,112, 2, 4, 5, 8, 9, 11 (n=7n = 7)

    • Minimum: 22

    • First Quartile (Q1Q_1): 22

    • Median (Q2Q_2): 55

    • Third Quartile (Q3Q_3): 99

    • Maximum: 1111

Boxplot of Data showing minimum of 2, Q1 of 2, median of 5, Q3 of 9, and maximum of 11
  • Example 2 Dataset: 2,3,4,5,5,5,5,8,8,10,11,122, 3, 4, 5, 5, 5, 5, 8, 8, 10, 11, 12 (n=12n = 12)

    • Minimum: 22

    • First Quartile (Q1Q_1): 4.254.25 (or ranked interpolation 44)

    • Median (Q2Q_2): 55

    • Third Quartile (Q3Q_3): 9.59.5

    • Maximum: 1212

Boxplot of Data2 showing minimum of 2, median of 5, and maximum of 12

Standard Error of the Mean

  • Standard Error Concept and Definition:

    • Standard error of the mean (abbreviated s.e.\text{s.e.} or sXˉs_{\bar{X}}) is the standard deviation of the sampling distribution of the sample mean.

    • While sample standard deviation (ss) quantifies the dispersion of individual raw data values around the sample mean, standard error quantifies how precisely the sample mean (Xˉ\bar{X}) estimates the true population mean (μ\mu).

  • Mathematical Formulation:

    • Calculated by dividing the sample standard deviation (ss) by the square root of the sample size (nn):

s.e.=sXˉ=sns.e. = s_{\bar{X}} = \frac{s}{\sqrt{n}}

  • Sample Size Relationship:

    • As sample size nn increases, the denominator n\sqrt{n} grows larger, which reduces the magnitude of the standard error.

    • This demonstrates that larger sample sizes yield sample means that are less variable, more stable, and more precise estimators of the population mean.

  • Standard Error Demonstration Dataset (n=24n = 24 observations):

    • Observations arranged in a 6×46 \times 4 matrix:

    • Row 1: 10,50,40,1010, 50, 40, 10

    • Row 2: 20,40,30,3020, 40, 30, 30

    • Row 3: 30,60,60,2030, 60, 60, 20

    • Row 4: 10,30,10,4010, 30, 10, 40

    • Row 5: 20,20,30,2020, 20, 30, 20

    • Row 6: 40,30,70,3040, 30, 70, 30

    • Key Statistical Takeaway: The standard deviation of the mean (standard error) is systematically smaller than the standard deviation of the original underlying observations, confirming that averages are inherently less variable than individual measurements.