Chapter 2: Data Description – Essential Concepts and Applications

Data Description

Understanding Cases and Variables

When analyzing data, it is crucial to identify the cases (the subjects or units being studied) and the variables (the characteristics measured on each case). Variables can be classified as categorical (qualitative) or quantitative (numerical).

• Categorical variables: Describe qualities or categories (e.g., gender, color, type).

• Quantitative variables: Represent numerical measurements (e.g., height, age, income).

• For categorical variables, identify the possible categories.

• For quantitative variables, note the units and measurement methods.

It is also important to determine whether the data come from an observational study or an experiment, and whether randomization was used.

Summary of Notation

Statistical notation helps succinctly describe populations and samples. The following table summarizes common symbols:

Symbol	Short Name	Full Description
p	Population proportion	Proportion of cases in a category in the population
\hat{p}	Sample proportion	Proportion of cases in a category in the sample
p_1 - p_2	Difference in two population proportions	Difference between proportions in two independent populations
\hat{p}_1 - \hat{p}_2	Difference in two sample proportions	Difference between proportions in two independent samples
\mu	Population mean	Mean of a quantitative variable in the population
\bar{x}	Sample mean	Mean of a quantitative variable in the sample
\mu_1 - \mu_2	Difference in two population means	Difference between means in two independent populations
\bar{x}_1 - \bar{x}_2	Difference in two sample means	Difference between means in two independent samples
\sigma	Population standard deviation	Standard deviation in the population
s	Sample standard deviation	Standard deviation in the sample
\rho	Population correlation	Linear correlation coefficient for the population
r	Sample correlation	Linear correlation coefficient for the sample
\alpha	Intercept of population LSR line	Intercept of least squares regression line for the population
a	Intercept of sample LSR line	Intercept of least squares regression line for the sample
\beta	Slope of population LSR line	Slope of least squares regression line for the population
b	Slope of sample LSR line	Slope of least squares regression line for the sample

Describing Distributions

The distribution of a variable shows the possible values and how often they occur. Distributions can be summarized and visualized in several ways, depending on the type of variable.

One Categorical Variable

• Summarized using a frequency table (counts) or relative frequency table (proportions).

• Visualized with a bar chart, where bar heights represent category proportions.

Two Categorical Variables

• Summarized with a two-way table (contingency table).

• Visualized with a segmented bar chart.

• Proportions can be calculated overall, by column, or by row to assess association.

One Quantitative Variable

• Describe the shape (skewed right, skewed left, bell-shaped, symmetric), center (mean, median), variability (standard deviation, IQR), and outliers.

• Visualized with histograms, dotplots, and box plots (based on the five-number summary).

Five-Number Summary and Box Plot

• The five-number summary consists of: minimum, first quartile (Q1), median (Q2), third quartile (Q3), and maximum.

• A box plot visually displays the five-number summary and highlights outliers.

• Outliers are plotted separately, and whiskers extend to the highest/lowest non-outlier values.

Measures of Center

• Mean (\bar{x}): Arithmetic average; sensitive to outliers.

• Median: Middle value; resistant to outliers.

• Use the mean for symmetric distributions without outliers; use the median for skewed distributions or those with outliers.

Measures of Variability

• Standard deviation (s): Measures typical distance from the mean; not resistant to outliers.

• Interquartile range (IQR): $IQR = Q_3 - Q_1$; measures spread of the middle 50% of values; resistant to outliers.

Outliers

• Values that are unusually high or low compared to the rest of the data.

• Use the 1.5 × IQR rule:

  • High outlier: $x > Q_3 + 1.5 \times IQR$

  • Low outlier: $x < Q_1 - 1.5 \times IQR$

The 95% Rule for Bell-Shaped Distributions

• In a bell-shaped (normal) distribution, about 95% of values are within two standard deviations of the mean:

$\text{Interval: } (\bar{x} - 2s,\ \bar{x} + 2s)$

• If the actual proportion within this interval is much different from 95%, the distribution is not bell-shaped.

• Estimate standard deviation: $\text{SD} \approx \frac{\text{width of middle 95%}}{4}$

Z-Score

• The z-score for a value $x$ is:

$z = \frac{x - \bar{x}}{s}$

• Indicates how many standard deviations a value is from the mean.

• Higher absolute z-scores indicate more unusual values.

Associations Between Variables

Two Categorical Variables

• Summarize with a two-way table.

• Assess association by comparing proportions across categories.

• If proportions are similar across rows/columns, there is likely no association.

One Categorical and One Quantitative Variable

• Compare typical values (mean or median) of the quantitative variable across categories.

• Large differences suggest association; small differences suggest little or no association.

Two Quantitative Variables

• Describe association in terms of strength, direction, and form (linear or not).

• Visualize with a scatter plot:

  • Horizontal axis: explanatory variable

  • Vertical axis: response variable

• Draw lines at the means to divide the plot into quadrants; association is present if points cluster in opposite quadrants.

Correlation Coefficient (r)

• Measures strength and direction of linear association:

• Range: $-1 \leq r \leq 1$

• Interpretation:

  • Strong: $|r| > 0.7$

  • Moderate: $0.5 < |r| \leq 0.7$

  • Weak: $0.3 < |r| \leq 0.5$

• Positive r: as one variable increases, so does the other.

• Negative r: as one variable increases, the other decreases.

Regression

A regression line models the linear relationship between two quantitative variables. The equation is:

$\hat{y} = a + b x$

• $\hat{y}$: predicted response

• $a$: intercept (predicted value when $x = 0$)

• $b$: slope (change in predicted response for a one-unit increase in $x$)

The residual for a data point $(x, y)$ is:

$\text{residual} = y - \hat{y}$

• Positive residual: observed value is above the regression line.

• Negative residual: observed value is below the regression line.

• The intercept may not always be meaningful, especially if $x = 0$ is outside the observed data range.

Examples and Applications

• Example (Categorical): A survey records whether people have a college education (yes/no) and whether they agree with a statement. A two-way table can show the proportion of each group that agrees or disagrees.

• Example (Quantitative): Heights of students are summarized with a five-number summary and visualized with a box plot to identify outliers.

• Example (Regression): Predicting weight from height using a regression line; the slope tells how much weight is expected to increase per inch of height.

Additional info: This summary includes expanded definitions, formulas, and examples to provide a self-contained study guide for introductory statistics students.