Analyzing Categorical Data and Graphical Representations

Fundamentals of Categorical and Quantitative Variables

• Quantitative Variable:

  • Definition: A variable that takes numerical values representing a measured or counted quantity.
  • Examples include measured continuous or discrete amounts such as height, weight, time, house counts, or total individual responses.

• Categorical Variable:

  • Definition: A variable that takes on values consisting of category names or group labels.
  • Examples include categorical classifications such as presidential administrations, neighborhood names, decades, or mascot choices.

• Graphical Displays for Categorical Data:

  • Bar Graphs: Displays categorical distributions where each category is represented by a separate bar whose height corresponds to either frequency or relative frequency.
  • Pie Charts: Circular charts divided into sectors that demonstrate relative frequencies (percentages) of each category relative to the whole.
  • Frequency: The total literal count of individuals or items belonging to a specific category.
  • Relative Frequency: The proportion or percentage of individuals or items falling into a specific category relative to the complete total sample size.

Two-Way Tables and Categorical Proportions

• Structure of Two-Way Tables:

  • Used to organize, summarize, and compare data involving two categorical variables simultaneously.
  • Consists of rows representing categories of one variable and columns representing categories of a second variable, complete with marginal total counts along the edges and a grand total count.

• Calculating Percentages and Proportions from Two-Way Tables:

  • Marginal Percentage:
    • Definition: The percentage distribution of a single categorical variable relative to the overall total sample.
    • Formula/Relationship: $\text{Marginal \%} = \frac{\text{Row Total or Column Total}}{\text{Grand Total}} = \frac{B}{C}$
    • Where $B$ represents a given row or column total, and $C$ represents the grand total.
  • Joint Percentage:
    • Definition: The percentage of individuals or items that satisfy conditions from two distinct categories simultaneously (the intersection of a row and a column).
    • Formula/Relationship: $\text{Joint \%} = \frac{\text{Cell Count}}{\text{Grand Total}} = \frac{A}{C}$
    • Where $A$ represents a specific cell count, and $C$ represents the grand total.
  • Conditional Percentage:
    • Definition: The percentage of individuals or items within a specific restricted subgroup (a single row or column) that satisfy a second category condition.
    • Formula/Relationship: $\text{Conditional \%} = \frac{\text{Cell Count}}{\text{Subgroup Row or Column Total}} = \frac{A}{B}$
    • Where $A$ represents the specific cell count, and $B$ represents the row or column total for that restricted subgroup.

Identifying Misleading Graphical Representations

• Vertical Axis Baseline Standard:

  • The vertical axis of any bar graph displaying frequencies or relative frequencies must begin precisely at $0$ (zero).
  • Starting a vertical axis at a non-zero value (truncating the axis) visually distorts category differences, exaggerating minor numerical variances into large visual differences.

• Avoidance of Pictographs and Images:

  • Using pictorial icons or images in place of standard rectangular bars in bar graphs creates misleading visual representations.
  • When scaling images to represent larger quantities, both height and width expand simultaneously, causing the image's overall area to scale quadratically rather than linearly, distorting visual perception.

Case Study: Unemployment Rates Across Presidencies

• Data Summary:

  • Unemployment rate at the end of presidency for President A: $7.8\%$
  • Unemployment rate at the end of presidency for President B: $5.4\%$
  • Unemployment rate at the end of presidency for President C: $4.0\%$

• Evaluation of Misleading Visual Representation:

  • The initial graph utilized a truncated vertical axis that began at $4\%$ rather than $0\%$ (with scale increments showing $4\%$, $5.4\%$, $6\%$, $7.8\%$, and $8\%$).
  • This non-zero baseline created an exaggerated display where the bar for President C (4.0\%$) appeared to have no height at all, while the bar for President A (7.8\%$) appeared disproportionately larger than President C, even though the actual difference was only $3.8$ percentage points ($7.8\% - 4.0\% = 3.8\%$).

• Requirements for Corrected Visual Representation:

  • The vertical axis must be re-anchored to start at $0\%$.
  • The vertical axis scale should ascend uniformly (e.g., $0\%$, $2\%$, $4\%$, $6\%$, $8\%$).
  • With a $0\%$ baseline, the heights of the bars accurately communicate the true proportional relationship between President A (7.8\%$), President B (5.4\%$), and President C ($4.0\%$).

Case Study: Housing Construction Analysis Across Neighborhoods and Decades

• Dataset Overview:

  • A real estate agent collected data tracking the number of houses constructed across three neighborhoods (Shady Lane, Oakcrest, Pinewood Estates) across three decades (1960s, 1970s, 1980s).

• Complete Two-Way Frequency Table:

  • 1960s:
    • Shady Lane: $40$
    • Oakcrest: $60$
    • Pinewood Estates: $0$
    • 1960s Row Total: $100$
  • 1970s:
    • Shady Lane: $30$
    • Oakcrest: $15$
    • Pinewood Estates: $45$
    • 1970s Row Total: $90$
  • 1980s:
    • Shady Lane: $10$
    • Oakcrest: $5$
    • Pinewood Estates: $15$
    • 1980s Row Total: $30$
  • Marginal Column Totals (Neighborhood Totals):
    • Shady Lane Total: $40 + 30 + 10 = 80$
    • Oakcrest Total: $60 + 15 + 5 = 80$
    • Pinewood Estates Total: $0 + 45 + 15 = 60$
    • Overall Grand Total: $80 + 80 + 60 = 220$ (or $100 + 90 + 30 = 220$)

• Statistical Computations and Analysis:

  • Proportion of Houses Built in Pinewood Estates:
    • Marginal proportion formula: $\frac{\text{Total Pinewood Estates Houses}}{\text{Grand Total}}$
    • Calculation: $\frac{60}{220} = \frac{3}{11} \approx 0.2727$
  • Distribution of Decade Built (Relative Frequencies):
    • 1960s Relative Frequency: $\frac{100}{220} = \frac{5}{11} \approx 45.5\%$
    • 1970s Relative Frequency: $\frac{90}{220} = \frac{9}{22} \approx 40.9\%$
    • 1980s Relative Frequency: $\frac{30}{220} = \frac{3}{22} \approx 13.6\%$
    • Total Relative Frequency Sum: $45.5\% + 40.9\% + 13.6\% = 100.0\%$
  • Joint Percentage of Homes Built in Oakcrest AND in the 1960s:
    • Joint percentage formula: $\frac{\text{Oakcrest and 1960s Cell Count}}{\text{Grand Total}}$
    • Calculation: $\frac{60}{220} = \frac{3}{11} \approx 27.3\%$

Case Study: High School Mascot Selection, Segmented Bar Graphs, and Mosaic Plots

• Survey Background:

  • In 1969, a newly constructed high school needed to select a mascot.
  • The principal surveyed random samples of teachers and students to vote between three mascot options: Rams, Falcons, or Prairie Dogs.

• Sample Relative Frequency Distributions:

  • Teacher Sample Relative Frequencies:
    • Rams: $80\%$ ($0.80$)
    • Falcons: $10\%$ ($0.10$)
    • Prairie Dogs: $10\%$ ($0.10$)
    • Total Teacher Sample: $100\%$
  • Student Sample Relative Frequencies:
    • Rams: $30\%$ ($0.30$)
    • Falcons: $60\%$ ($0.60$)
    • Prairie Dogs: $10\%$ ($0.10$)
    • Total Student Sample: $100\%$

• Graphical Representations:

  • Side-by-Side Bar Graphs: Displays separate relative frequency bar charts for Teachers and Students on a scale ranging from $0.0\%$ to $100.0\%$.
  • Segmented Bar Graph:
    • Combines relative frequencies by stacking segments vertically within a single bar of total height $100.0\%$ for each group.
    • Teacher Stack: $80\%$ Rams segment at base, $10\%$ Falcons segment in middle, $10\%$ Prairie Dogs segment at top.
    • Student Stack: $30\%$ Rams segment at base, $60\%$ Falcons segment in middle, $10\%$ Prairie Dogs segment at top.

• Initial Support Interpretation (Relative Frequency Only):

  • Looking strictly at relative frequency percentage bars, Rams initially appears to have the most support because the individual relative frequency bar for Rams among teachers (80\%$) is the largest single segment.\n\n* **Sample Size Disparity and Count Calculations**:\n * The survey sampled 100$teachers and$500 students.\n * Calculated Count Table:\n * **Teacher Sample Counts (N = 100)**:\n * Rams: 0.80 \times 100 = 80 teachers\n * Falcons: 0.10 \times 100 = 10 teachers\n * Prairie Dogs: 0.10 \times 100 = 10 teachers\n * Total Teachers Sampled = 100\n * **Student Sample Counts (N = 500)**:\n * Rams: 0.30 \times 500 = 150 students\n * Falcons: 0.60 \times 500 = 300 students\n * Prairie Dogs: 0.10 \times 500 = 50 students\n * Total Students Sampled = 500\n\n* **Transition from Segmented Bar Graph to Mosaic Plot**:\n * **Sample Size Ratio**: \frac{500 \text{ students}}{100 \text{ teachers}} = 5 times more students were sampled than teachers.\n * **Mosaic Plot Construction**:\n * A Mosaic Plot is created by modifying a segmented bar graph so that the **width** of each stack is proportional to the total sample size of that group.\n * To adjust the graph for sample sizes, the Student bar is drawn 5 times wider than the Teacher bar.\n\n* **Final Choice Determination**:\n * **Absolute Support Counts**:\n * Rams Total Votes: 80 \text{ (teachers)} + 150 \text{ (students)} = 230 total votes\n * Falcons Total Votes: 10 \text{ (teachers)} + 300 \text{ (students)} = 310 total votes\n * Prairie Dogs Total Votes: 10 \text{ (teachers)} + 50 \text{ (students)} = 60 total votes\n * **Mascot Recommendation and Explanation**:\n * The school should select **Falcons** as the official mascot.\n * Explanation: Although Rams had higher percentage support among teachers, Falcons received significantly more total votes (310$total votes compared to$230$for Rams and$60$$ for Prairie Dogs) due to strong support among the larger student sample size.