Week 8: Descriptive Statistics, Frequency Distributions, and the Normal Distribution

Classification of Variables in Medical Statistics

  • Taxonomy of Variables:

    • Variables represent measurable characteristics, clinical parameters, or demographic traits that differ among study participants.

    • Variables are broadly divided into two primary statistical classes: numeric (quantitative) and categorical (qualitative).

  • Numeric (Quantitative) Variables:

    • Continuous Numeric Variables: Variables that can assume any real numeric value along a continuous scale within a given range, restricted only by the precision of the measuring instrument.

    • Examples: Body weight (e.g., 51.0 kg51.0\,\text{kg}, 63.1 kg63.1\,\text{kg}, 83.0 kg83.0\,\text{kg}), standing height (e.g., 159 cm159\,\text{cm}, 167 cm167\,\text{cm}, 167.5 cm167.5\,\text{cm}), body mass index (BMI\text{BMI}, e.g., 20.17 kg/m220.17\,\text{kg/m}^2, 22.63 kg/m222.63\,\text{kg/m}^2, 29.58 kg/m229.58\,\text{kg/m}^2), and serum cholesterol levels in peripheral blood.

    • Discrete Numeric Variables: Variables that take distinct, separate numeric values, usually whole integers arising from count processes.

    • Examples: Number of cigarettes smoked per day, number of residents in a hotel, parity (number of live births: 11, 22, 33, 44), age at menarche (1111, 1212, 1313, 1414, 15 years15\,\text{years}), age at first full-term pregnancy (2222, 2323, 2525, 2727, 37 years37\,\text{years}), oral contraceptive duration in months (00, 1414, 2626, 5050, 6262, 86 months86\,\text{months}), and breastfeeding duration in months (00, 11, 33, 55, 66, 77, 88, 99, 1010, 1313, 15 months15\,\text{months}).

  • Categorical (Qualitative) Variables:

    • Normal Categorical (Nominal) Variables: Qualitative variables with two or more unordered, mutually exclusive categories that lack any natural rank, scale, or hierarchy.

    • Examples: Nationality (e.g., Cypriot, Romanian, Bulgarian, Russian, Lebanese), marital status (e.g., single, married, divorced, widowed), and case-control classification (e.g., case, control).

    • Binary (Dichotomous) Variables: A specific subtype of nominal variable with exactly two mutually exclusive outcomes.

    • Examples: Biological sex (female, male), ever use of oral contraceptives (yes, no), and ever breastfed (yes, no).

    • Ordered Categorical (Ordinal) Variables: Qualitative variables whose categories exhibit an intrinsic, natural order or clinical ranking, though the mathematical differences between adjoining categories cannot be presumed uniform.

    • Examples: Categorized BMI\text{BMI} (underweight, normal, overweight, obese) and level of motor disability (light, light-to-moderate, moderate, moderate-to-severe, severe).

  • Case-Control Breast Cancer Dataset Overview:

Measures of Central Tendency

  • Core Principle of Central Tendency:

    • Measures of central tendency identify a single central, typical, or representative value located in the middle of a probability or frequency distribution around which other measured values cluster.

  • The Mode:

    • Definition: The specific value that appears with the highest frequency in a given sample or distribution; represents the single most probable or typical expected outcome.

    • Example: In a survey quantifying the number of alcoholic drinks consumed during a social night out across a student cohort: 3,3,3,3,3,3,0,1,1,1,1,2,2,2,3,4,4,5,5,63, 3, 3, 3, 3, 3, 0, 1, 1, 1, 1, 2, 2, 2, 3, 4, 4, 5, 5, 6.

    • Frequency count for 00: 11

    • Frequency count for 11: 44

    • Frequency count for 22: 33

    • Frequency count for 33: 77

    • Frequency count for 44: 22

    • Frequency count for 55: 22

    • Frequency count for 66: 11

    • The modal value is 33 drinks, representing the single most common response.

    • Scientific Role: Rarely utilised as the primary descriptive summary statistic in scientific, epidemiological, or medical research due to its sensitivity to binning choices in continuous data and lack of mathematical algebraic manipulability.

  • The Median (MM):

    • Definition: The geographic middle point of an ordered distribution; exactly 50%50\% of observations have a value smaller than or equal to the median, and exactly 50%50\% have a value larger than or equal to the median.

    • Algorithmic Computation:

    1. Rank-order all nn individual observations from smallest to largest value.

    2. For an odd number of observations (nn is odd), the median corresponds to the value occupying the central position:        Position=n+12\text{Position} = \frac{n + 1}{2}

    3. For an even number of observations (nn is even), no single observation occupies the absolute center. The median is computed as the arithmetic mean of the two middle observations situated at positions n2\frac{n}{2} and n2+1\frac{n}{2} + 1:        Median=x(n/2)+x(n/2+1)2\text{Median} = \frac{x_{(n/2)} + x_{(n/2 + 1)}}{2}

    • Clinical Example: In a sample of n=101n = 101 patients with ordered systolic blood pressure recordings:

    • Because n=101n = 101 is an odd integer, the median lies at the 101+12=51st\frac{101 + 1}{2} = 51\text{st} ordered position.

    • The measured value at the 51st51\text{st} position is 122 mmHg122\,\text{mmHg}.

    • Thus, Median=122 mmHg\text{Median} = 122\,\text{mmHg}, meaning 50%50\% of patients exhibit blood pressure ≤122 mmHg\le 122\,\text{mmHg} and 50%50\% exhibit blood pressure ≥122 mmHg\ge 122\,\text{mmHg}.

Bar graph of 101 ordered blood pressure measurements highlighting the 51st observation as the median
  • The Arithmetic Mean (xˉ\bar{x}):

    • Definition: Commonly referred to as the average; calculated by summing all individual values in the sample and dividing by the total count of observations (nn):     xˉ=∑i=1nxin\bar{x} = \frac{\sum_{i=1}^{n} x_i}{n}

    • Qualities: Serves as the mathematical balance point (centre of gravity) of the data.

    • In a perfectly symmetrical (normal) distribution, the arithmetic mean coincides identically with the median and the mode:     xˉ=Median=Mode\bar{x} = \text{Median} = \text{Mode}

  • Worked Numerical Example for Central Tendency:

    • Raw dataset (n=9n = 9): 13,18,13,14,13,16,14,21,1313, 18, 13, 14, 13, 16, 14, 21, 13

    • Ordered dataset: 13,13,13,13,14,14,16,18,2113, 13, 13, 13, 14, 14, 16, 18, 21

    • Mode: The value 1313 occurs 44 times (higher frequency than any other value):     Mode=13\text{Mode} = 13

    • Median: For n=9n = 9, the position is 9+12=5th\frac{9 + 1}{2} = 5\text{th} value:     Median=14\text{Median} = 14

    • Mean: Summing all 99 observations and dividing by 99:     xˉ=13+13+13+13+14+14+16+18+219=1359=15\bar{x} = \frac{13 + 13 + 13 + 13 + 14 + 14 + 16 + 18 + 21}{9} = \frac{135}{9} = 15

    • Range: Difference between the extreme observations:     Range=21−13=8\text{Range} = 21 - 13 = 8

Measures of Dispersion

  • Core Principle of Dispersion:

    • Measures of dispersion describe the extent of spread, scatter, or variability of data values around their central location.

  • The Range:

    • Formula:     Range=xmax⁡−xmin⁡\text{Range} = x_{\max} - x_{\min}

    • Limitations: Extremely sensitive to sample size and heavily distorted by outliers. Because it relies exclusively on the two most extreme observations, it disregards the distribution of all data points between the minimum and maximum.

  • Quantiles, Quartiles, and Percentiles:

    • Quantiles are statistical division points calculated after sorting observations from lowest to highest:

    • Tertiles: Divide ordered observations into 33 equal segments (33.3%33.3\% of data each).

    • Quartiles: Divide ordered observations into 44 equal segments (25%25\% of data each).

    • Quintiles: Divide ordered observations into 55 equal segments (20%20\% of data each).

    • Percentiles: Divide ordered observations into 100100 equal segments (1%1\% of data each).

    • Standard Quartiles:

    • Q1Q_1 (25th25\text{th} percentile): The value below which 25%25\% of observations fall.

    • Q2Q_2 (50th50\text{th} percentile / Median): The value below which 50%50\% of observations fall.

    • Q3Q_3 (75th75\text{th} percentile): The value below which 75%75\% of observations fall.

  • The Interquartile Range (IQR\text{IQR}):

    • Formula:     IQR=Q3−Q1\text{IQR} = Q_3 - Q_1

    • Interpretation: The mathematical width of the central interval encompassing the middle 50%50\% of all ordered data.

    • Robustness: Because the 25%25\% lowest values and the 25%25\% highest values are ignored, the IQR\text{IQR} is resistant to extreme values, making it the preferred measure of dispersion for skewed distributions.

  • Variance (s2s^2) and Standard Deviation (ss):

    • Standard Deviation (ss): Quantifies the average or typical spread of individual observations around the arithmetic mean (xˉ\bar{x}).

    • Sequential Computation:

    1. Compute the sample mean (xˉ\bar{x}).

    2. Calculate individual deviations of each observation from the mean:        Deviationi=xi−xˉ\text{Deviation}_i = x_i - \bar{x}

    3. Square each deviation to prevent positive and negative values from canceling each other out:        Squared Deviationi=(xi−xˉ)2\text{Squared Deviation}_i = (x_i - \bar{x})^2

    4. Sum the squared deviations and divide by the degrees of freedom (n−1n - 1) to produce sample variance (s2s^2):        s2=∑i=1n(xi−xˉ)2n−1s^2 = \frac{\sum_{i=1}^{n} (x_i - \bar{x})^2}{n - 1}

    5. Calculate the positive square root of variance to yield the standard deviation (ss), returning the metric to the original units of measurement:        s=s2=∑i=1n(xi−xˉ)2n−1s = \sqrt{s^2} = \sqrt{\frac{\sum_{i=1}^{n} (x_i - \bar{x})^2}{n - 1}}

Scatter plot with deviations from the mean alongside a vertical bar graph showing signed deviation magnitudes
  • When multiple datasets share the same mean, differing standard deviations illustrate varying degrees of spread around that centre:

Histograms and Distribution Shapes

  • The Histogram:

    • Graphical tool for summarizing the distribution of a single numeric variable.

    • Built by dividing the continuous range of variable values into equal-sized, contiguous, non-overlapping intervals termed "bins" or class intervals.

    • The vertical axis displays the frequency (absolute count) or relative frequency of observations contained within each bin.

  • Histogram Construction Example (Age Dataset):

  • Interpreting Patterns in Histograms: - Frequency distribution

    • When describing a distribution, three characteristics are evaluated: shape, centre, and spread.

    • Directly connecting bin midpoints produces jagged, distracting detail; fitting a continuous smoothed curve reveals the underlying distribution shape.

Histogram with fitted smooth curve highlighting the general distribution shape
  • Modality of Distributions:

    • Unimodal Distribution: Characterised by a single prominent frequency peak.

    • Bimodal Distribution: Characterized by two prominent, distinct peaks. A bimodal pattern often indicates that the dataset is a mixture of two separate underlying populations.

    • Example: Total adult body weight distributions combining adult males and females often produce two distinct peaks—an average female peak at 166 lbs166\,\text{lbs} and an average male peak at 196 lbs196\,\text{lbs}.

Bimodal distribution of average human body weight showing female and male peaks

Distribution Symmetry, Skewness, and Outliers

  • Types of Distribution Symmetry:

    • Symmetrical (Normal or Bell-Shaped / Gaussian): The left and right halves around the central value are approximate mirror images. Skewness coefficient is zero (Skewness=0\text{Skewness} = 0).

    • Positively Skewed (Skewed to the Right): The right tail extends further out than the left tail, pulled by high values. Skewness coefficient is positive (Skewness>0\text{Skewness} > 0).

    • Negatively Skewed (Skewed to the Left): The left tail extends further out than the right tail, pulled by low values. Skewness coefficient is negative (Skewness<0\text{Skewness} < 0).

Comparison of curves showing negatively skewed, normal, and positively skewed distributions
  • Mathematical Impact of Skewness on Central Tendency:

    • In a Normal Distribution: The mean, median, and mode are identical and sit directly in the middle:     xˉ=Median=Mode\bar{x} = \text{Median} = \text{Mode}

    • In a Positively Skewed Distribution: Extreme values in the right tail pull the arithmetic mean upward, while the mode stays at the distribution's peak and the median remains between the mode and mean:     Mode<Median<xˉ\text{Mode} < \text{Median} < \bar{x}

    • In a Negatively Skewed Distribution: Extreme values in the left tail pull the arithmetic mean downward:     xˉ<Median<Mode\bar{x} < \text{Median} < \text{Mode}

Three curves displaying the relative order of mean, median, and mode across skewed and symmetric distributions
  • Comparative Clinical Survival Example:

    • Stomach Cancer Survival Times (Symmetrical):

    • Mean: xˉ=3.4 years\bar{x} = 3.4\,\text{years}

    • Median: M=3.4 yearsM = 3.4\,\text{years}

    • Symmetrical distribution where the mean and median match.

    • Multiple Myeloma Survival Times (Positively Skewed):

    • Mean: xˉ=3.4 years\bar{x} = 3.4\,\text{years}

    • Median: M=2.5 yearsM = 2.5\,\text{years}

    • The long survival times of a few patients pull the mean upward (3.4 years3.4\,\text{years}) relative to the median (2.5 years2.5\,\text{years}).

Histograms comparing symmetrical survival in stomach cancer versus positively skewed survival in multiple myeloma
  • Outliers:

    • Definition: Observations that fall well outside the overall pattern of the distribution, separated by distinct gaps from the main cluster of data.

    • Analytical Principle: Outliers must be identified and examined for clinical and biological plausibility rather than automatically discarded.

    • Plausibility Example (Physical Activity):

    • Main distribution: 88 to 16 hours/month16\,\text{hours/month}.

    • Extreme data points: 5−6 hours/month5 - 6\,\text{hours/month} (very low activity) and 18−19 hours/month18 - 19\,\text{hours/month} (very high activity).

    • Because these observations represent realistic human behavior, they are retained in the analysis.

Histogram of monthly physical activity hours highlighting low and high activity outliers
  • Impact of Outliers on Mean Versus Median:

    • In a study tracking daily cigarettes smoked:

    • Dataset without outliers: xˉ=3.4\bar{x} = 3.4, M=3.4M = 3.4.

    • Dataset with high-end outliers added (e.g., 13−14 cigarettes/day13 - 14\,\text{cigarettes/day}):

      • The mean is pulled substantially to the right from 3.43.4 to 4.24.2.

      • The median shifts only slightly from 3.43.4 to 3.63.6, demonstrating resistance to extreme values.

The Boxplot and Five-Number Summary

  • Structure of a Boxplot:

    • A graphical display of continuous distributions based on the five-number summary and statistical fences:

    • Central Box: Stretches from Q1Q_1 (25th25\text{th} percentile) to Q3Q_3 (75th75\text{th} percentile), enclosing the middle 50%50\% of observations (IQR\text{IQR}).

    • Internal Line: Marks the median (Q2Q_2, 50th50\text{th} percentile).

    • Fences: Mathematical thresholds used to detect potential outliers:       Upper Fence=Q3+1.5×IQR\text{Upper Fence} = Q_3 + 1.5 \times \text{IQR}       Lower Fence=Q1−1.5×IQR\text{Lower Fence} = Q_1 - 1.5 \times \text{IQR}

    • Whiskers: Extend from the box to the most extreme data points within the upper and lower fences.

    • Outlier Points: Data values that fall beyond the statistical fences, plotted as individual circles.

  • Worked Boxplot Example (Disease X Survival Data):

Boxplot diagram illustrating the five-number summary, fences, and outlier threshold for Disease X
  • Identifying Distribution Shape from Boxplots:

    • Symmetric (Normal): The median line divides the box roughly in half, and the whiskers are equal in length.

    • Positively Skewed: The median line is shifted toward the lower quartile (Q1Q_1), the upper whisker is notably longer than the lower whisker, and outliers appear past the upper fence.

    • Negatively Skewed: The median line is shifted toward the upper quartile (Q3Q_3), the lower whisker is notably longer than the upper whisker, and outliers appear past the lower fence.

Boxplots shown alongside their corresponding normal, right-skewed, and left-skewed distribution curves

Selecting Summary Statistics Based on Distribution Shape

  • Selection Criteria:

    • The distribution of numeric variables should always be inspected using a histogram, a boxplot, or both prior to selecting descriptive summary statistics.

  • Decision Guidelines:

    • Symmetrical (Normal) Distributions Without Outliers:

    • Measure of Central Tendency: Arithmetic Mean (xˉ\bar{x}).

    • Measures of Dispersion: Standard Deviation (ss) and Variance (s2s^2).

    • Skewed Distributions or Data with Outliers:

    • Measure of Central Tendency: Median (MM).

    • Measures of Dispersion: Interquartile Range (IQR\text{IQR}) and Quartiles (Q1,Q3Q_1, Q_3).

  • Effect of Sample Size:

    • In large samples (n>500n > 500 individuals), a small number of outliers will have a negligible effect on the arithmetic mean, permitting its use.

    • Skewness affects the arithmetic mean regardless of sample size; a large sample does not correct or offset distribution asymmetry.

The Normal (Gaussian) Distribution and the Empirical Rule

  • Theoretical Foundation:

    • A continuous probability distribution characterized by a symmetric, bell-shaped curve.

    • Defined entirely by two parameters: the mean (μ\mu), which sets the center and peak, and the standard deviation (σ\sigma), which determines the width and spread.

    • The total area under the probability density curve equals 1.01.0 (100%100\% probability).

    • Varying the standard deviation alters the curve's height and spread without changing the total area under the curve:

Three normal distribution curves with identical mean of 50 and standard deviations of 5, 10, and 20
  • Physical Weight Stack Analogy:

Gym weight stack wear pattern compared side by side with a normal distribution curve
  • The Empirical Rule (68–95–99.7% Rule):

    • For any normally distributed variable:

      • Approximately 68%68\% (precisely 68.27%68.27\%) of all observations fall within 11 standard deviation of the mean:       μ±1σ\mu \pm 1\sigma

      • Approximately 95%95\% (precisely 95.45%95.45\%) of all observations fall within 22 standard deviations of the mean:       μ±2σ\mu \pm 2\sigma

      • Exactly 95.00%95.00\% of observations lie within 1.961.96 standard deviations of the mean:       μ±1.96σ\mu \pm 1.96\sigma

      • Approximately 99.7%99.7\% (precisely 99.73%99.73\%) of all observations fall within 33 standard deviations of the mean:       μ±3σ\mu \pm 3\sigma

Normal distribution curve marked with 68 percent, 95 percent, and 99.7 percent standard deviation intervals
  • Clinical and Biological Applications of the Empirical Rule:

    • Standing Height in a Large Sample:

    • Parameters: Mean μ=1.774 m\mu = 1.774\,\text{m}, Standard Deviation σ=0.1465 m≈0.147 m\sigma = 0.1465\,\text{m} \approx 0.147\,\text{m}.

    • 68%68\% Interval (μ±1σ\mu \pm 1\sigma):       μ−1σ=1.774−0.147=1.627 m\mu - 1\sigma = 1.774 - 0.147 = 1.627\,\text{m}       μ+1σ=1.774+0.147=1.921 m≈1.92 m\mu + 1\sigma = 1.774 + 0.147 = 1.921\,\text{m} \approx 1.92\,\text{m}       About 68%68\% of individuals have heights between 1.63 m1.63\,\text{m} and 1.92 m1.92\,\text{m}.

    • 95%95\% Interval (μ±1.96σ\mu \pm 1.96\sigma):       1.96×0.1465=0.286 m1.96 \times 0.1465 = 0.286\,\text{m}       μ−1.96σ=1.774−0.286=1.488 m≈1.49 m\mu - 1.96\sigma = 1.774 - 0.286 = 1.488\,\text{m} \approx 1.49\,\text{m}       μ+1.96σ=1.774+0.286=2.060 m=2.06 m\mu + 1.96\sigma = 1.774 + 0.286 = 2.060\,\text{m} = 2.06\,\text{m}       About 95%95\% of individuals have heights between 1.49 m1.49\,\text{m} and 2.06 m2.06\,\text{m}.

    • Systolic Blood Pressure (n=276n = 276 Participants):

    • Parameters: Mean μ=128.5 mmHg\mu = 128.5\,\text{mmHg}, Standard Deviation σ=15.2 mmHg\sigma = 15.2\,\text{mmHg}.

    • Expected 68%68\% range:       128.5±15.2 mmHg→[113.3 mmHg, 143.7 mmHg]128.5 \pm 15.2\,\text{mmHg} \rightarrow [113.3\,\text{mmHg},\, 143.7\,\text{mmHg}]

    • Systolic Blood Pressure Reference Example:

    • Parameters: Mean μ=120 mmHg\mu = 120\,\text{mmHg}, Standard Deviation σ=15 mmHg\sigma = 15\,\text{mmHg}.

    • 68%68\% of values fall within 120±15 mmHg120 \pm 15\,\text{mmHg} (105105 to 135 mmHg135\,\text{mmHg}).

    • 95%95\% of values fall within 120±30 mmHg120 \pm 30\,\text{mmHg} (9090 to 150 mmHg150\,\text{mmHg}).

    • 99.7%99.7\% of values fall within 120±45 mmHg120 \pm 45\,\text{mmHg} (7575 to 165 mmHg165\,\text{mmHg}).

Questions & Discussion

  • Question 1: Probabilities from a Normal Height Distribution:

    • Prompt: Given a normal distribution of height with μ=1.774 m\mu = 1.774\,\text{m} and σ=0.1465 m\sigma = 0.1465\,\text{m}:

    1. What is the probability that a randomly chosen individual has a height <1.49 m< 1.49\,\text{m}?

    2. What is the probability that a randomly chosen individual has a height >1.92 m> 1.92\,\text{m}?

    • Response & Derivations:

    • For part 1: The cutoff 1.49 m1.49\,\text{m} equals μ−1.96σ\mu - 1.96\sigma (1.774−0.286=1.488 m≈1.49 m1.774 - 0.286 = 1.488\,\text{m} \approx 1.49\,\text{m}). The central range μ±1.96σ\mu \pm 1.96\sigma contains 95%95\% of the distribution, leaving 5%5\% split equally between both tails (2.5%2.5\% in each tail). Therefore:       P(Height<1.49 m)=2.5%=0.025P(\text{Height} < 1.49\,\text{m}) = 2.5\% = 0.025

    • For part 2: The cutoff 1.92 m1.92\,\text{m} equals μ+1σ\mu + 1\sigma (1.774+0.147=1.921 m≈1.92 m1.774 + 0.147 = 1.921\,\text{m} \approx 1.92\,\text{m}). The central range μ±1σ\mu \pm 1\sigma contains approximately 68%68\% of the distribution, leaving 32%32\% in the tails (16%16\% in each tail). Therefore:       P(Height>1.92 m)=16%=0.16P(\text{Height} > 1.92\,\text{m}) = 16\% = 0.16

  • Question 2: Expected Number of Underweight Individuals in a Cohort:

    • Prompt: In a cohort of n=600n = 600 individuals, Body Mass Index (BMI\text{BMI}) follows a normal distribution. Clinicians define as "underweight" anyone whose BMI\text{BMI} is more than 22 standard deviations below the population mean (<μ−2σ< \mu - 2\sigma). How many individuals are expected to meet this definition?

    • Response & Derivation:

    • By the Empirical Rule, μ±2σ\mu \pm 2\sigma covers approximately 95%95\% of the distribution.

    • The remaining 5%5\% of the probability area is divided equally between the upper and lower tails (2.5%2.5\% each):       P(BMI<μ−2σ)=2.5%=0.025P(\text{BMI} < \mu - 2\sigma) = 2.5\% = 0.025

    • Multiplying by the sample size n=600n = 600:       600×0.025=15 individuals600 \times 0.025 = 15\,\text{individuals}

  • Question 3: Evaluating Blood Pressure Variability Between Two Clinical Groups:

    • Prompt: Systolic blood pressure is recorded in two groups of patients:

    • Group A: Mean=125 mmHg\text{Mean} = 125\,\text{mmHg}, SD=5 mmHg\text{SD} = 5\,\text{mmHg}

    • Group B: Mean=125 mmHg\text{Mean} = 125\,\text{mmHg}, SD=20 mmHg\text{SD} = 20\,\text{mmHg}

    • Which statement is the most accurate?

      • A. Group A has a higher average blood pressure than Group B.

      • B. Group B has greater variability in blood pressure measurements.

      • C. Most patients in Group B have higher blood pressure than patients in Group A.

      • D. Group A has four times as many patients as Group B.

      • E. The distributions of both groups must be normally distributed.

    • Response & Rationale:

    • Correct Answer: B.

    • Both cohorts have an identical mean systolic blood pressure (125 mmHg125\,\text{mmHg}). However, Group B's standard deviation (20 mmHg20\,\text{mmHg}) is four times larger than Group A's (5 mmHg5\,\text{mmHg}), reflecting greater dispersion and variability in blood pressure measurements.

  • Question 4a: Distribution Shape of Hospital Stay Duration:

    • Prompt: A histogram displays the length of hospital stay (days) among a group of patients:

    • Bin 0−5 days0 - 5\,\text{days}: Frequency =25= 25

    • Bin 5−10 days5 - 10\,\text{days}: Frequency =12= 12

    • Bin 10−15 days10 - 15\,\text{days}: Frequency =6= 6

    • Bin 15−20 days15 - 20\,\text{days}: Frequency =3= 3

    • Bin 25−30 days25 - 30\,\text{days}: Frequency =1= 1

Histogram of length of hospital stay in days showing right-side tail extension
* Which statement is the most appropriate?
  * A. The distribution is approximately normal.
  * B. The mean is likely to be lower than the median.
  * C. The distribution is positively skewed.
  * D. The distribution is negatively skewed.
  * E. The mean and median must be equal.
  • Response & Rationale:

    • Correct Answer: C.

    • The histogram shows a peak in the lowest bin (0−5 days0 - 5\,\text{days}) and a long tail extending toward higher numbers of days on the right, which defines a positively skewed (right-skewed) distribution.

    • Question 4b: Selecting a Measure of Dispersion for Skewed Hospital Stay Data:

  • Prompt: Based on the positively skewed hospital stay distribution, which measure of dispersion is most appropriate for summarizing the data?

    • A. Standard deviation

    • B. Mean

    • C. Interquartile range

    • D. Range

    • E. Variance

  • Response & Rationale:

    • Correct Answer: C.

    • For skewed distributions, the interquartile range (IQR\text{IQR}) is preferred over the standard deviation, variance, or range because it is resistant to skewness and extreme values in the tails.