9/22/26:Discrete Random Variables and Probability Distribution Functions

Foundations of Discrete Random Variables and Probability Distributions

  • Probability Measure Overview:

    • Probability provides a quantitative measure of how likely an event is to occur.
    • Values always lie on a closed interval between 0%0\% and 100%100\%, or as a decimal between 00 and 11 inclusive (0≤P≤10 \le P \le 1).
    • Classical probability is computed by defining the sample space and taking the ratio of successful outcomes to total possible outcomes:         Probability=Number of Successful OutcomesTotal Number of Sample Space Outcomes\text{Probability} = \frac{\text{Number of Successful Outcomes}}{\text{Total Number of Sample Space Outcomes}}
  • Probability Distribution Function (PDF):

    • In statistics, a PDF is a mathematical function that describes the theoretical or empirical probability distribution over all possible outcomes of a random variable.
    • A PDF is not a digital document file format; it is a probability distribution function.
  • Discrete Random Variables (DRVs):

    • Variable: Represents a statistical inquiry or question asked of a population.
    • Random Variable: A variable whose numerical outcomes are determined purely by chance.
    • Discrete: Characterized by distinct, separate values with finite or countable gaps between them. Continuous values (such as fractional parts or decimals) cannot occur between discrete increments.
  • Examples of Discrete vs. Non-Discrete Quantities:

    • Coin Flipping: Asking a coin whether it lands on heads or tails results in a random outcome.
    • Card Drawing: Drawing the top card from a deck yields a suit (heart, spade, diamond, or club) randomly, without deterministic temporal patterns.
    • Sibling Counts: Countable values such as 00, 11, 22, 33, or 44. Outcomes like 2.52.5 or 2.752.75 siblings are impossible.
    • Die Rolling: Outcomes are strictly bounded to integer faces {1,2,3,4,5,6}\left\{1, 2, 3, 4, 5, 6\right\}. Fractional rolls like 3.53.5 or 6.56.5 cannot exist.
  • Notation:

    • X\mathbf{X} (Capital Letter): Denotes the random variable itself (the statistical question or process).
    • x\mathbf{x} (Lowercase Letter): Denotes a specific numerical value that the random variable XX can attain.

Classroom Coin Flipping Experiment and Frequency Distributions

  • Experimental Protocol:

    • Each participant flips a single coin (nickel) 1010 times.
    • The random variable XX represents the total number of heads obtained out of 1010 flips (n=10n = 10).
  • Classroom Experimental Data Collection:

    • Total sample size of participants reporting data: N=22N = 22 trials.
    • Recorded counts of heads (xx) across participants included values such as 7,6,4,4,7,7,3,6,8,8,3,37, 6, 4, 4, 7, 7, 3, 6, 8, 8, 3, 3.
    • Summarized Frequency Distribution across N=22N = 22 total responses:
      • x=3x = 3 heads: Frequency f=2f = 2, Relative Frequency = 222=111≈9.11%\frac{2}{22} = \frac{1}{11} \approx 9.11\%
      • x=4x = 4 heads: Frequency f=2f = 2, Relative Frequency = 222≈9.1%\frac{2}{22} \approx 9.1\%
      • x=5x = 5 heads: Frequency f=5f = 5, Relative Frequency = 522≈22.7%\frac{5}{22} \approx 22.7\%
      • x=6x = 6 heads: Frequency f=7f = 7, Relative Frequency = 722≈31.8%\frac{7}{22} \approx 31.8\%
      • x=7x = 7 heads: Frequency f=4f = 4, Relative Frequency = 422≈18.2%\frac{4}{22} \approx 18.2\%
      • x=8x = 8 heads: Frequency f=2f = 2, Relative Frequency = 222≈9.1%\frac{2}{22} \approx 9.1\%
  • Experimental Analysis:

    • Mode / Most Frequent Outcome: x=6x = 6 heads occurred most frequently (31.8%31.8\% relative frequency).
    • Possible Range for 1010 Flips: x∈{0,1,2,3,4,5,6,7,8,9,10}x \in \left\{0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10\right\}.
    • Possible Range for 1,0001,000 Flips: x∈{0,1,2,…,1000}x \in \left\{0, 1, 2, \dots, 1000\right\}.
      • Even with 1,0001,000 flips, XX remains a discrete random variable because non-integer outcomes (e.g., 99.599.5 heads) are strictly impossible.
    • Expected Range for 1,0001,000 Flips: For a fair coin, theoretical expected heads is 500500. Practical observations across large populations typically cluster in the range of 500500 to 700700 heads.
    • Sum of Relative Frequencies: Adding all relative frequencies yields exactly 100%100\% (or 1.001.00 in decimal form):         9.11%+9.1%+22.7%+31.8%+18.2%+9.1%=100%9.11\% + 9.1\% + 22.7\% + 31.8\% + 18.2\% + 9.1\% = 100\%

Axioms and Mathematical Properties of Probability Distribution Functions (PDFs)

  • Definition of a Probability Distribution Function (PDF):

    • A function that assigns a probability P(x)P(x) to each possible value xx of a discrete random variable XX.
  • Two Axiomatic Properties of any valid PDF:

    1. Bounded Probability Property: Every individual probability output must be non-negative and no greater than 11:         0≤P(x)≤1for all x0 \le P(x) \le 1 \quad \text{for all } x
    2. Completeness Property: The sum of probabilities over all possible outcomes in the sample space must equal exactly 11 (or 100%100\%):         ∑P(x)=1\sum P(x) = 1
  • Connection to Descriptive Statistics:

    • A PDF table operates identically to a relative frequency table.
    • Frequency tables derived from empirical sampling transition into theoretical probability distribution functions as sample sizes become infinitely large.

Case Study 1: Store Cookie Search Distribution

  • Scenario:

    • Random variable XX represents the number of grocery stores driven to before finding a specific cookie brand.
    • Finding cookies at each store represents a set of mutually exclusive events (once cookies are purchased at store xx, searching terminates).
  • PDF Table:

    • x=1x = 1 store: P(1)=0.55P(1) = 0.55 (55%55\%
    • x=2x = 2 stores: P(2)=0.20P(2) = 0.20 (20%20\%
    • x=3x = 3 stores: P(3)=0.15P(3) = 0.15 (15%15\%
    • x=4x = 4 stores: P(4)=0.10P(4) = 0.10 (10%10\%
  • Verification of PDF Axioms:

    • All individual probabilities satisfy 0≤P(x)≤10 \le P(x) \le 1
    • Sum of probabilities:         ∑P(x)=0.55+0.20+0.15+0.10=1.00(100%\sum P(x) = 0.55 + 0.20 + 0.15 + 0.10 = 1.00 \quad (100\%
    • It is 100%100\% certain that finding the cookies will require driving to either 11, 22, 33, or 44 stores.

Case Study 2: Blueberry Count Analysis per Pint

  • Scenario:

    • Sarah samples pint containers of blueberries at a grocery store and records the count of blueberries (XX) per pint.
  • Empirical Probability Distribution Table:

    • x=147x = 147 blueberries: P(147)=0.01P(147) = 0.01 (1%1\%
    • x=148x = 148 blueberries: P(148)=0.02P(148) = 0.02 (2%2\%
    • x=149x = 149 blueberries: P(149)=0.05P(149) = 0.05 (5%5\%
    • x=150x = 150 blueberries: P(150)=0.08P(150) = 0.08 (8%8\%
    • x=151x = 151 blueberries: P(151)=0.18P(151) = 0.18 (18%18\%
    • x=152x = 152 blueberries: P(152)=0.34P(152) = 0.34 (34%34\%
    • x=153…156x = 153 \dots 156 blueberries: Remaining probability mass distributed up to 156156 blueberries.
  • Question 1: Mode (Most Common Count):

    • x=152x = 152 blueberries is the mode, as it exhibits the highest individual probability (P(152)=0.34P(152) = 0.34 or 34%34\%
  • Question 2: Probability of At Least 150 Blueberries (P(X≥150)P(X \ge 150)):

    • Direct Addition Method (utilizing mutual exclusivity):         P(X≥150)=P(150)+P(151)+P(152)+P(153)+P(154)+P(155)+P(156)=0.92(92%P(X \ge 150) = P(150) + P(151) + P(152) + P(153) + P(154) + P(155) + P(156) = 0.92 \quad (92\%
    • Complement Rule Method:         P(X≥150)=1−P(X<150)P(X \ge 150) = 1 - P(X < 150)P(X<150)=P(147)+P(148)+P(149)=0.01+0.02+0.05=0.08P(X < 150) = P(147) + P(148) + P(149) = 0.01 + 0.02 + 0.05 = 0.08P(X≥150)=1−0.08=0.92(92%P(X \ge 150) = 1 - 0.08 = 0.92 \quad (92\%
  • Question 3: Sample Mean (xˉ\bar{x} or Expected Value μ\mu):

    • Formula for the mean of a PDF table:         xˉ=∑[x⋅P(x)]\bar{x} = \sum [x \cdot P(x)]
    • Step-by-step calculation:         xˉ=(147⋅0.01)+(148⋅0.02)+(149⋅0.05)+⋯=151.94\bar{x} = (147 \cdot 0.01) + (148 \cdot 0.02) + (149 \cdot 0.05) + \dots = 151.94
    • Interpretation: Across repeated sampling, Sarah finds an average of approximately 151.94151.94 blueberries per pint container. The sample mean must logically lie between the distribution minimum (147147) and maximum (156156).
  • Question 4: Sample Median:

    • The median represents the point at which cumulative probability reaches or crosses 50%50\% (0.500.50).
    • Cumulative calculation:
      • Up through x=151x = 151: 1%+2%+5%+8%+18%=34%1\% + 2\% + 5\% + 8\% + 18\% = 34\% cumulative probability.
      • At x=152x = 152: Cumulative probability reaches 34%+34%=68%34\% + 34\% = 68\%.
    • The 50%50\% threshold is crossed at x=152x = 152, making 152152 blueberries the median.

Expected Value and Standard Deviation Formulas

  • Expected Value (E[X]E[X] or μ\mu):

    • The theoretical long-term average outcome resulting from repeating an experiment infinitely many times.
    • Under the Law of Large Numbers, as the number of trials approaches infinity, the empirical average converges to the theoretical expected value.
    • Formula:         μ=E[X]=∑[x⋅P(x)]\mu = E[X] = \sum [x \cdot P(x)]
  • Variance (σ2\sigma^2) and Standard Deviation (σ\sigma):

    • Measures the spread or dispersion of outcomes relative to the mean μ\mu.
    • The standard deviation serves as a statistical ruler measuring distance from the center.
    • Formula for Variance:         σ2=∑[(x−μ)2⋅P(x)]\sigma^2 = \sum \left[(x - \mu)^2 \cdot P(x)\right]
    • Formula for Standard Deviation:         σ=∑[(x−μ)2⋅P(x)]\sigma = \sqrt{\sum \left[(x - \mu)^2 \cdot P(x)\right]}

Example Walkthrough: Flipping Four Fair Coins

  • Experiment Parameters:

    • A fair coin is flipped 44 times (n=4n = 4).
    • Random variable XX = number of heads obtained (x∈{0,1,2,3,4}x \in \left\{0, 1, 2, 3, 4\right\}).
  • Given PDF Table:

    • x=0x = 0: P(0)=0.0625P(0) = 0.0625
    • x=1x = 1: P(1)=0.2500P(1) = 0.2500
    • x=2x = 2: P(2)=0.3750P(2) = 0.3750
    • x=3x = 3: P(3)=0.2500P(3) = 0.2500
    • x=4x = 4: P(4)=0.0625P(4) = 0.0625
  • Step 1: Expected Value (μ\mu) Calculation:     μ=∑[x⋅P(x)]\mu = \sum [x \cdot P(x)]μ=(0⋅0.0625)+(1⋅0.2500)+(2⋅0.3750)+(3⋅0.2500)+(4⋅0.0625)\mu = (0 \cdot 0.0625) + (1 \cdot 0.2500) + (2 \cdot 0.3750) + (3 \cdot 0.2500) + (4 \cdot 0.0625)μ=0+0.25+0.75+0.75+0.25=2.0\mu = 0 + 0.25 + 0.75 + 0.75 + 0.25 = 2.0

    • Interpretation: Over long-term repetition of flipping 44 coins, an average of 22 heads is expected.
  • Step 2: Standard Deviation (σ\sigma) Calculation Table:

    • For x=0x = 0: (0−2)2⋅0.0625=(−2)2⋅0.0625=4⋅0.0625=0.25(0 - 2)^2 \cdot 0.0625 = (-2)^2 \cdot 0.0625 = 4 \cdot 0.0625 = 0.25
    • For x=1x = 1: (1−2)2⋅0.2500=(−1)2⋅0.2500=1⋅0.2500=0.25(1 - 2)^2 \cdot 0.2500 = (-1)^2 \cdot 0.2500 = 1 \cdot 0.2500 = 0.25
    • For x=2x = 2: (2−2)2⋅0.3750=(0)2⋅0.3750=0⋅0.3750=0.00(2 - 2)^2 \cdot 0.3750 = (0)^2 \cdot 0.3750 = 0 \cdot 0.3750 = 0.00
    • For x=3x = 3: (3−2)2⋅0.2500=(1)2⋅0.2500=1⋅0.2500=0.25(3 - 2)^2 \cdot 0.2500 = (1)^2 \cdot 0.2500 = 1 \cdot 0.2500 = 0.25
    • For x=4x = 4: (4−2)2⋅0.0625=(2)2⋅0.0625=4⋅0.0625=0.25(4 - 2)^2 \cdot 0.0625 = (2)^2 \cdot 0.0625 = 4 \cdot 0.0625 = 0.25
    • Sum of weighted squared deviations:         ∑[(x−μ)2⋅P(x)]=0.25+0.25+0.00+0.25+0.25=1.00\sum \left[(x - \mu)^2 \cdot P(x)\right] = 0.25 + 0.25 + 0.00 + 0.25 + 0.25 = 1.00
    • Take the square root to calculate standard deviation:         σ=1.00=1.0\sigma = \sqrt{1.00} = 1.0
  • Empirical Rule Application and Z-Scores:

    • Within 11 standard deviation of the mean (μ±1σ→2±1→[1,3]\mu \pm 1\sigma \rightarrow 2 \pm 1 \rightarrow [1, 3]):         P(1≤X≤3)=P(1)+P(2)+P(3)=0.25+0.375+0.25=0.875(87.5%P(1 \le X \le 3) = P(1) + P(2) + P(3) = 0.25 + 0.375 + 0.25 = 0.875 \quad (87.5\%
    • Within 22 standard deviations of the mean (μ±2σ→2±2→[0,4]\mu \pm 2\sigma \rightarrow 2 \pm 2 \rightarrow [0, 4]):         P(0≤X≤4)=1.00(100%P(0 \le X \le 4) = 1.00 \quad (100\%
    • Z-score: Quantifies the exact number of standard deviations a specific data value xx lies away from the mean μ\mu:         z=x−μσz = \frac{x - \mu}{\sigma}

Questions & Classroom Discussion

  • Bimodal Distributions:

    • Question: What happens if two different xx values share the same highest frequency?
    • Response: Either value can be reported, or both values are stated. The distribution is classified as bimodal.
  • Identifying Central Tendency:

    • Question: Which standard measure of central tendency corresponds to the most frequently occurring value?
    • Response: The mode.
  • Median Determination on Cutoff Boundaries:

    • Question: How is the median reported if a percentile threshold lands exactly on a discrete outcome boundary?
    • Response: If a discrete outcome boundary falls exactly on the 50th50\text{th} percentile mark, any value between that boundary number and the next discrete data value is technically valid. However, if the 50th50\text{th} percentile falls strictly within a specific data interval, that precise discrete value must be chosen.