Describing Data Numerically
Numerical Summaries and the Analytical Framework
Curricular Progression and Context:
- Graphical and tabular summaries provide initial visual structures for categorical and quantitative distributions.
- Numerical descriptive statistics quantify essential distributional properties, specifically central tendency, variability, and relative position.
- These univariate measures establish the analytical foundation required to evaluate bivariate quantitative relationships.
Primary Divisions of Numerical Description:
- Measures of Center: Mean, median, and mode.
- Measures of Variability: Range, variance, and standard deviation.
- Grouped and Weighted Calculations: Weighted mean.
- Measures of Relative Position and Outlier Detection: -scores, percentiles, quartiles, and the interquartile range ().
- Five-Number Summary and Boxplots: Order statistics, exploratory data analysis, and fence calculations.
Measures of Central Tendency
The Mean:
- The mean is the arithmetic average and the most common measure of center.
- Mathematical Notation:
- denotes the total number of observations in a population.
- denotes the total number of observations in a sample.
- represents an observed variable value.
- (capital Greek sigma) denotes summation across all observations.
- Population Mean ():
- Represented by the Greek letter mu ():
- Sample Mean ():
- Represented by -bar ():
- Computational procedure:
- Calculate the sum of all observed data values ().
- Divide the resulting sum by the total sample size ().
Physical Interpretation as the Balance Point and Outlier Sensitivity:
- The mean acts as the physical fulcrum or balance point of a distribution.
- An extreme outlier heavily shifts the mean toward the tail, whereas the median remains stable.

Step-by-Step Applications of the Mean:
- Example 3.1: Calculating the Sample Mean:
- Data set (hours of nightly sleep reported by students): .
- Sample size: .
- Sum of observations:
- Sample mean:
- Example 3.2: Determining an Unknown Observation from a Given Mean:
- Total songs purchased: .
- Target mean price: .
- Price of four individual songs: each.
- Total expenditure required:
- Cost of the four known songs:
- Price of the fifth song ():
- Target Mean Exam Problem:
- An academic course contains exams. The first four percentage scores are , , , and .
- Required mean for the course: .
- Total points required:
- Cumulative points obtained on exams 1 through 4:
- Required score on the fifth exam ():
The Median:
- The median is the physical midpoint of a data set arranged in ascending numerical order.
- Exactly of observations lie at or below the median, and lie at or above it.
- Resistance Property: The median is robust and resistant to extreme values and outliers because its value depends solely on rank order rather than numerical magnitude.
- Operational Algorithm:
- Arrange all observations in non-decreasing order.
- If is odd, the median is the unique single middle value located at position:
- If is even, the median is the arithmetic mean of the two middle values located at positions:
- Example 3.5: Finding the Median for Odd and Even Sample Sizes:
- Sample A (Student absences):
- Arranged data (, odd):
- Position:
- Sample B (Student absences):
- Arranged data (, even):
- Middle positions: and , corresponding to values and .
- Median calculation:
The Mode:
- The mode is the observation that appears with the highest frequency in a data set.
- A data set can have a single mode, multiple modes, or no mode.
- Modal Classifications:
- No Mode: Occurs when every value appears with identical frequency (e.g., all values occur exactly once).
- Unimodal: Exactly one value displays the greatest frequency.
- Bimodal: Exactly two distinct values share the identical maximum frequency.
- Multimodal: More than two distinct values share the identical maximum frequency.
- Stating the category "bimodal" does not identify the modes; the specific numerical values must be identified.
- Example 3.6: Identifying Modes:
- Distribution A (Phone calls per hour):
- Frequency of 7 is ; frequency of 6 is ; all other values appear once.
- (Distribution is unimodal).
- Distribution B (Phone calls per hour):
- Every value occurs exactly once.
- Result: No mode.
- Distribution C (Phone calls per hour):
- Each value appears with equal frequency ().
- Result: No mode (neither unimodal nor bimodal).
Comparative Case Study: Music Videos:
- Counts of music videos by artist:
- Michael Jackson:
- Taylor Swift:
- Usher:
- Katy Perry:
- Sample mean ():
- Sample median:
- Ordered values:
- Sample mode: Value appears twice.
Worked Example: Retirement Ages:
- Sample data ():
- Ordered array:
- Mean calculation:
- Median calculation:
- Mode calculation: appears twice.
Comparative Case Study: Hank Aaron vs. Barry Bonds Home Runs:
- Hank Aaron Career Data (23 seasons):
- Data:
- Total home runs:
- Mean:
- Median ( ordered position):
- Mode: (appears times).
- Barry Bonds Career Data (22 seasons):
- Data:
- Total home runs:
- Mean:
- Median (average of and positions, both ):
- Mode: Multimodal with modes (each appears twice).
- Analytical Comparison:
- Both players exhibit an identical median performance of home runs.
- Bonds has a higher mean ( vs. ), influenced by the extreme upper value of .
- Aaron shows consistency at high productivity, with a single mode of .
- One cannot definitively conclude superior overall performance without considering variability and outlier effects.
Distribution Shape and Measure Selection
- Geometric Relationships Between Mean, Median, and Mode:
- Mode: The horizontal coordinate corresponding to the absolute peak (maximum height) of the density curve.
- Median: The line that divides the total area under the density curve into two equal halves ( each).
- Mean: Pulled away from the peak toward the tail containing extreme values or skewness.


- Rules for Skewness and Central Tendency:
- Symmetric Unimodal Distribution:
- Right-Skewed (Positively Skewed) Distribution:
- Left-Skewed (Negatively Skewed) Distribution:

Graphical Deduction Example:
- In a right-skewed unimodal distribution with vertical marker lines labeled , , and :
- Line represents the peak: Mode.
- Line splits the area in half: Median.
- Line is pulled furthest into the long right tail: Mean.
Application to Skewed Economic Data:
- The distribution of individual annual incomes in the United States is strongly skewed to the right.
- In 2004, the income values for the top were reported as and .
- Because right-skewness pulls the mean above the median:
Criteria for Selecting the Appropriate Measure of Center:
- Qualitative (Categorical) Data: The mode is the only valid measure of center.
- Quantitative Data without Outliers / Approximately Symmetric: The mean is preferred because it incorporates the magnitude of every observation.
- Quantitative Data with Outliers or Heavy Skewness: The median is preferred because of its resistance to extreme values.
Example 3.8: Evaluating Specific Data Scenarios:
- Women's T-shirt sizes (S, M, L, XL): Categorical/ordinal data; use the Mode.
- Professional baseball team player salaries: Highly right-skewed quantitative distribution with extreme high outliers; use the Median.
- Home sales prices in a subdivision of similar tract homes: Homogeneous quantitative data without extreme outliers; use the Mean.
- Professor evaluation ratings (Best, Average, Worst): Categorical/ordinal ranking; use the Mode.
Summary Properties of Central Tendency Measures:
| Property | Mean | Median | Mode |
|---|---|---|---|
| Common Description | Arithmetic Average | Middle Value | Most Frequent Value |
| Must Belong to Data Set? | No | No (can be average of middle pair) | Yes |
| Number of Possible Values | Exactly one unique value | Exactly one unique value | One, multiple, or none |
| Sensitivity to Outliers | Highly sensitive | Highly resistant (robust) | Highly resistant (robust) |
| Optimal Data Domain | Quantitative, symmetric, no outliers | Quantitative, skewed, or with outliers | Qualitative / Categorical |
Measures of Variability
- The Significance of Variation:
- Measures of center fail to describe the spread of a distribution.
- Two distinct distributions can have identical means, medians, and modes while possessing completely different dispersion profiles.
- Case Study: Women's Volleyball Team Heights (Inches):
- Western Massachusetts University (WMU):
- , ,
- Northern Connecticut University (NCU):
- , ,

The Range:
- The range is the difference between the maximum and minimum observations in a data set:
- Calculation for volleyball teams:
- WMU Range:
- NCU Range:
- Analytical Limitations: The range relies solely on the two most extreme values, ignores all intermediate data points, and is vulnerable to outliers.
The Concept of Deviation:
- A deviation quantifies the directed distance between an individual observation and the mean of the data set.
- Sample Deviation:
- Population Deviation:
- Directional Properties:
- If , the deviation is positive.
- If , the deviation is negative.
- If , the deviation equals zero.
- Fundamental Algebraic Property: The sum of all raw deviations around the mean always equals zero:
- Raw deviations cannot be averaged directly without squaring or applying absolute values.
Population Variance () and Standard Deviation ():
- Population Variance (): The arithmetic mean of the squared deviations from the population mean:
- Population Standard Deviation (): The positive square root of the population variance, restoring the original unit of measurement:
- Represents the typical distance an observation lies from the population mean.
Sample Variance () and Sample Standard Deviation ():
- Sample statistics serve as point estimators for population parameters.
- Dividing sample squared deviations by underestimates the population variance.
- Degrees of Freedom Correction: Dividing by corrects for this downward bias.
- Sample Variance ():
- Sample Standard Deviation ():
- represents the typical difference between an observation and the sample mean.
Computational (Shortcut) Formulas:
- Reduce rounding errors caused by intermediate decimal calculations.
- Population Parameters:
- Sample Statistics:
Worked Example: Dog Shoulder Heights:
- Heights of five dogs: .
- Sample size: .
- Mean Calculation:

- Deviations and Sum of Squared Deviations Table:
| Value () | Deviation () | Squared Deviation () |
|---|---|---|
| Sum (Total) |
Sample Variance ():
Sample Standard Deviation ():
An individual dog's height differs from the mean of by an average of .
- Worked Example: Daily Temperature Variations:
Daily recorded temperatures: ().
Sum of values:
Mean:
Deviations ():
Squared deviations:
Sum of squared deviations:
Sample variance:
Sample standard deviation:
- Fundamental Mathematical Properties of Standard Deviation:
Measures the spread of data about the mean; must only be used when the mean describes the center.
if and only if all observations are identical (zero variability). Under all other circumstances, .
As the dispersion around the mean increases, increases.
Standard deviation shares the vulnerability of the mean to extreme outliers.
- Practical Interpretations of Standard Deviation:
Example 3.13: Stock Price Volatility:
- Profacto Corporation daily closing price standard deviation:
- Yardsmoth Company daily closing price standard deviation:
- Interpretation: Profacto stock exhibits far tighter clustering around its average price, indicating lower volatility and risk for long-term investors. Yardsmoth exhibits wider swings, appealing to investors seeking rapid profit through high volatility.
Course Exam Section Comparison (Wayne State University STA 1020):
- Morning section exam score standard deviation:
- Evening section exam score standard deviation:
- Interpretation: Scores in the evening section are clustered much more tightly around their mean than morning scores. This statistic does not show which section achieved a higher average score, as variability does not measure performance level.
Grouped Data and the Weighted Mean
The Weighted Mean:
- Used when individual observations differ in relative importance or contribution to the final measure.
- Mathematical Formula:
- represents the value of observation .
- represents the numerical weight assigned to observation .
- Computational Steps:
- Multiply each observed value by its corresponding weight factor .
- Sum all products: .
- Divide this sum by the total weight: .
Step-by-Step Weighted Mean Problems:
- Walter's U.S. History Course Grade:
- Grading breakdown: Tests (), Homework (), Quizzes (), Final Exam ().
- Walter's scores: Tests = , Homework = , Quizzes = , Final Exam = .
- Calculation:
- Heather's Target Course Grade Problem:
- Grading breakdown: Tests (), Class Assignments (), Semester Project (), Final Exam ().
- Known scores: Tests = , Class Assignments = , Semester Project = .
- Target course grade: .
- Setting up the equation with final exam score :
- Heather must score on the final exam to attain a final grade of .
- Current Completed Course Average Problem:
- Grade components completed:
- Homework ( weight): Current average =
- Quizzes ( weight): Current average =
- Participation ( weight): Current average =
- Midterm Exams ( weight): Current average =
- Final Exam ( weight): Uncompleted
- Cumulative weight of completed course components:
- Weighted sum of completed points:
- Current course grade:
Measures of Relative Position and Outlier Detection
Standard Scores (-Scores):
- A -score indicates the number of standard deviations an individual observation falls above or below the mean.
- Standardizing values allows direct comparison between observations from different distributions with distinct means and standard deviations.
- Formulas:
- For a Sample:
- For a Population:
- Sign Conventions:
- Positive -score: Observation lies above the mean ().
- Negative -score: Observation lies below the mean ().
- Zero -score: Observation equals the mean ().
- Recovering the Raw Value () from a Given -Score:
- Sample:
- Population:
- If , , and :
Step-by-Step Applications of -Scores:
- Example 3.25: Daily Summer Temperature:
- Average summer daily high: , with .
- Temperature of interest: .
- Calculation:
- Standardized SAT Performance:
- Math SAT parameters: , .
- Jasmine's test score: .
- -score:

Comparative Performance: Emma vs. Sophia:
- Emma scored in a class with and :
- Sophia scored in a class with and :
- Comparison: Sophia performed better relative to her peer group ().
Comparative Performance: Maria vs. Elena (5K Footrace):
Maria ran in a race where and :
Elena ran in a race where and :
Comparison: In competitive racing, lower completion times indicate superior performance. Elena finished standard deviations below her race average, outperforming Maria relative to their respective fields.
Outlier Detection via the -Score Method:
An outlier is an extreme data point relative to the rest of the distribution.
Standard Outlier Classification Thresholds:
- Not Unusual: Observations where .
- Moderately Unusual: Observations where or .
- Outlier: Observations where or .
Customer Volume Problem:
- Normal daily volume: , . Observed day: .
- Standard score calculation:
- Conclusion: Because , serving customers is classified as moderately unusual.
Delivery Route Time Problem:
Route duration parameters: , . Driver completion time: .
Standard score calculation:
Conclusion: Because , completing the route in is an outlier.
Percentiles and Percentile Ranks:
Percentile Definition: Let be an integer such that . The percentile () is the value such that at least of all data values fall at or below it.
Location Algorithm for the Percentile:
- Sort data in ascending order.
- Compute the position index :
- Rule 1 (Fractional/Decimal Result): If contains a decimal, round up to the next integer. The observation at that integer rank is the percentile.
- Rule 2 (Integer Result): If is an exact integer, the percentile is the arithmetic mean of the value at position and the value at position :
Percentile Rank of an Observed Value:
The percentile rank corresponding to a specific value is calculated as:
The resulting value is rounded to the nearest integer.
Quartiles and the Interquartile Range ():
Quartiles divide an ordered distribution into four equal parts, each containing of the data.
- First Quartile (): The percentile.
- Second Quartile (): The percentile, identical to the Median.
- Third Quartile (): The percentile.

Interquartile Range ():
The difference between the third and first quartiles:
Measures the dispersion of the central of the distribution.
Highly resistant to extreme values and outliers.
Quartile Calculations for Motorist Speeds:
Sample A ():
- Median (): Average of () and () values:
- Lower half (7 observations: ):
- Upper half (7 observations: ):
Sample B ():
Median (): Located at position , which is .
Lower half (7 observations below position 8: ):
Upper half (7 observations above position 8: ):
Quartile and Analysis for State Exports Data ():
Ordered export values: .

First quartile (): Average of positions 3 and 4 ( and ):
Median (): Average of positions 6 and 7 ( and ):
Third quartile (): Average of positions 9 and 10 ( and ):
Interquartile Range ():
- Comprehensive Reference Study: Lake Water Pollutant Concentrations:
Ordered concentrations ():
Percentile:
- Since is fractional, round up to position .
- Observation at position 7:
Percentile:
- Since is an integer, average positions 14 and 15 ( and ):
Quartiles ():
- Position for ( percentile): . Average positions 10 and 11 ( and ):
- Position for ( percentile): . Average positions 20 and 21 ( and ):
- Position for ( percentile): . Average positions 30 and 31 ( and ):
Percentile Rank of :
- Values span positions 1 through 33 (count = ).
- Rank calculation:
Percentile Rank of :
- Values span positions 1 through 24 (count = ).
- Rank calculation:
Percentile Rank of :
- Values span positions 1 through 39 (count = ).
- Rank calculation:
The Five-Number Summary and Boxplots
The Five-Number Summary:
- A resistant summary of a quantitative distribution consisting of five order statistics:
- Minimum (Min): The smallest numerical observation in the data set.
- First Quartile (): The percentile.
- Median (): The percentile.
- Third Quartile (): The percentile.
- Maximum (Max): The largest numerical observation in the data set.
Detecting Outliers Using the Rule:
- The rule sets operational fences beyond which values are classified as outliers.
- Interquartile Range:
- Lower Fence (LF):
- Upper Fence (UF):
- Outlier Criterion: Any observation or is an outlier.
Step-by-Step Construction of a Boxplot (Box-and-Whisker Plot):
- Step 1: Construct a horizontal numerical scale spanning the data range, including the calculated fences.
- Step 2: Plot vertical line segments above the axis corresponding to , the median (), and .
- Step 3: Connect the segments at and to form a central rectangle (box) with the median line inside.
- Step 4: Draw horizontal whiskers extending from the ends of the box:
- Lower whisker extends from down to the smallest data value greater than or equal to the lower fence.
- Upper whisker extends from up to the largest data value less than or equal to the upper fence.
- Step 5: Plot any individual observations located outside the fences as isolated points (outliers).

Worked Example: Creating a Boxplot from a Five-Number Summary:
- Five-number summary: , , , , .
- Interquartile Range ():
- Fence Calculations:
- Lower Fence:
- Upper Fence:
- Outlier Evaluation:
- The minimum value () exceeds the lower fence (); no lower outlier exists.
- The maximum value () exceeds the upper fence (); is an outlier.
- In a modified boxplot, the upper whisker stops at the largest non-outlier value, and is marked with an isolated symbol.
Worked Example: Outlier Analysis for Lake Water Data ():
- Previously calculated values:
- Fence Calculations:
- Lower Fence:
- Upper Fence:
- Outlier Assessment:
- Data range: , .
- Since and , all observations fall within the fences.
- Result: The lake water pollutant distribution contains no outliers under the criterion.