Comprehensive Study Guide on Quartiles, Five-Number Summaries, Standard Deviation, and Calculator Procedures

Fundamentals of Quartiles and Medians

• Rope Folding Analogy for Data Splitting:

  • Folding a rope in half and cutting it creates two equal sections. This represents the median ($\text{Med}$ or $Q_2$), which divides a dataset into two equal halves (the lower $50\%$ and the upper $50\%$ of the data).
  • Taking the lower half of the rope, folding it in half, and cutting it yields the first quartile ($Q_1$). The first quartile is the median of the lower half of the dataset.
  • Taking the upper half of the rope, folding it in half, and cutting it yields the third quartile ($Q_3$). The third quartile is the median of the upper half of the dataset.
  • Quartiles are fundamentally medians of their respective halves of the dataset.

• Calculation Conventions:

  • The primary convention used in statistical textbooks and Quantitative Reasoning defines quartiles strictly as the medians of the lower and upper halves.
  • Alternative algorithms exist (such as those used in Microsoft Excel), but the median-based definition serves as the core convention.

• General Rules for Medians:

  • Odd Number of Observations ($n$ is odd): The median is the exact middle observation.
  • Even Number of Observations ($n$ is even): The median is the average of the two central observations.

• Splitting Datasets to Determine Quartiles:

  • Even Case:
  • When total observations $n$ is even, the dataset divides cleanly into two halves containing an equal number of observations ($n^*$).
  • Example: A dataset with $n = 6$ observations splits into a lower half of $n^* = 3$ and an upper half of $n^* = 3$.
  • Since $n^* = 3$ is odd, each quartile corresponds to exactly one observation in its respective half.
  • Quartile Location Formula:$\text{Location of Quartile} = \frac{n^* + 1}{2}$       where $n^*$ represents the number of observations in that specific half.
  • For $n^* = 3$:       $\text{Location} = \frac{3 + 1}{2} = 2$       This indicates that $Q_1$ is the $2\text{nd}$ observation in the lower half, and $Q_3$ is the $2\text{nd}$ observation in the upper half.
  • The relative location of $Q_1$ within the lower half must strictly match the relative location of $Q_3$ within the upper half.
  • Odd Case:
  • When total observations $n$ is odd, the overall median is a single exact data point (e.g., the $4\text{th}$ observation in a dataset of $n = 7$).
  • When splitting the dataset into halves to calculate quartiles, the single central median value is excluded/ignored.
  • Excluding the median ensures both halves contain an identical number of observations ($n^*$).
  • The central median value remains part of the Five-Number Summary; it is only omitted from the sub-calculations for $Q_1$ and $Q_3$.

Five-Number Summary and Stemplot Example: New England Patriots (Textbook 2.15)

• Dataset Context:

  • Subject: Weights of $10$ defensive linemen from the 2019–2020 roster of the New England Patriots (winners of the 2019 Super Bowl).
  • Raw observations consist of 3-digit integer weight values in pounds.

• Stemplot Construction Rules:

  • Stems represent the first two digits (hundreds and tens places).
  • Leaves represent the final digit (ones place) and must always be a single digit in length.
  • Data range: Minimum value = $250\,\text{lbs}$, Maximum value = $345\,\text{lbs}$.
  • Stems ordered from smallest to largest: $25$, $26$, $27$, $28$, $29$, $30$, $31$, $32$, $33$, $34$.
  • Stemplots order data sequentially, simplifying the identification of the Five-Number Summary.

• Calculating the Five-Number Summary ($n = 10$):

  • Minimum ($\text{Min}$):$\text{Min} = 250\,\text{lbs}$
  • Median ($\text{Med}$):
  • Since $n = 10$ (even), the median is the average of the $5\text{th}$ and $6\text{th}$ ordered values.     $\text{Location} = \frac{10 + 1}{2} = 5.5$
  • The $5\text{th}$ observation is $275\,\text{lbs}$ and the $6\text{th}$ observation is $280\,\text{lbs}$.     $\text{Median} = \frac{275 + 280}{2} = 277.5\,\text{lbs}$
  • Splitting the Data:
  • The split occurs directly between $275$ and $280$.
  • Lower half contains $n^* = 5$ observations; upper half contains $n^* = 5$ observations.
  • First Quartile ($Q_1$):
  • Lower half has $n^* = 5$ (odd).     $\text{Location} = \frac{5 + 1}{2} = 3$
  • $Q_1$ is the $3\text{rd}$ value in the lower half:     $Q_1 = 260\,\text{lbs}$
  • Third Quartile ($Q_3$):
  • $Q_3$ is the $3\text{rd}$ value in the upper half counting upward from the median split line:     $Q_3 = 300\,\text{lbs}$
  • Maximum ($\text{Max}$):$\text{Max} = 345\,\text{lbs}$
  • Full Five-Number Summary:$250\,\text{lbs}, \quad 260\,\text{lbs}, \quad 277.5\,\text{lbs}, \quad 300\,\text{lbs}, \quad 345\,\text{lbs}$
  • Numerical Validity Check: Always verify that the summary values progress monotonically from smallest to largest:     $\text{Min} \le Q_1 \le \text{Med} \le Q_3 \le \text{Max}$

Interquartile Range (IQR) and Outlier Identification

• Interquartile Range Definition and Formula:

  • The Interquartile Range ($\text{IQR}$) measures the variability/spread of the middle $50\%$ of the data.
  • It is defined as the distance between the third quartile and the first quartile.   $\text{IQR} = Q_3 - Q_1$

• Properties of the IQR:

  • Must always be non-negative:     $\text{IQR} \ge 0$
  • Definitional reason: Distance cannot be negative.
  • Mathematical reason: $Q_3 \ge Q_1$, so $Q_3 - Q_1 \ge 0$.

• Note on Box Plots:

  • Box-and-whisker plots rely on the $\text{IQR}$ to construct whiskers. In formal statistical analyses, box plots are historically used to check normality assumptions, but superior statistical tests exist today. Stemplots remain preferable for analyzing distribution shape.

• The $1.5 \times \text{IQR}$ Outlier Rule:

  • Establishes a mathematical boundary interval. Any data point falling strictly outside this interval is flagged as a suspected outlier.   $\text{Lower Bound} = Q_1 - 1.5 \times \text{IQR}$$\text{Upper Bound} = Q_3 + 1.5 \times \text{IQR}$
  • Analogy: Works similarly to reference range values on medical blood test reports.

• Example: New York Travel Times (Textbook 2.8):

  • Dataset: $n = 20$ travel times (in minutes) for New York workers, including a long travel time of $75\,\text{minutes}$.
  • Five-Number Summary for New York Travel Times:
  • $\text{Min} = 10\,\text{minutes}$
  • $\text{Med} = \frac{30 + 35}{2} = 32.5\,\text{minutes}$ ($10\text{th}$ and $11\text{th}$ values)
  • $Q_1 = \frac{12 + 15}{2} = 13.5\,\text{minutes}$ ($5\text{th}$ and $6\text{th}$ values of lower $10$)
  • $Q_3 = \frac{50 + 50}{2} = 50\,\text{minutes}$ ($5\text{th}$ and $6\text{th}$ values of upper $10$)
  • $\text{Max} = 75\,\text{minutes}$
  • Calculating $\text{IQR}$:$\text{IQR} = 50 - 13.5 = 36.5\,\text{minutes}$
  • Calculating Outlier Boundaries:$\text{Lower Bound} = 13.5 - 1.5 \times 36.5 = 13.5 - 54.75 = -41.25\,\text{minutes}$$\text{Upper Bound} = 50 + 1.5 \times 36.5 = 50 + 54.75 = 104.75\,\text{minutes}$
  • Order of Operations Requirement: Multiplication ($1.5 \times \text{IQR}$) must be computed before performing addition or subtraction.
  • Outlier Evaluation:
  • Check if any original observation is $< -41.25$ or $> 104.75$.
  • The value $75\,\text{minutes}$ falls well within the upper bound of $104.75\,\text{minutes}$.
  • Conclusion: There are no suspected outliers in this dataset.

Standard Deviation and Measures of Spread

• Definition of Standard Deviation ($s$ or $s_x$):

  • The most widely used measure of variability in statistics.
  • Measures the spread/dispersion of individual data values around the sample mean ($\bar{x}$).
  • Higher dispersion from the mean results in a larger standard deviation.

• Degrees of Freedom ($df$):

  • Definitional meaning: Represents the number of observations that can vary freely. If $n$ total values are known alongside summary statistics, $n-1$ values can vary freely while the final value is fixed.
  • Statistical meaning: Reflects sample size adequacy and capacity to perform statistical inference.

• Properties of Standard Deviation:

  1. Non-negativity:$s \ge 0$
  • Definitional reason: It represents a physical distance from the mean.
  • Mathematical reason: Deviations $(x_i - \bar{x})$ are squared (making all terms positive), summed, divided by $n-1$, and square-rooted.
  1. Sensitivity to Outliers:
  • Standard deviation is non-resistant and highly sensitive to extreme values. An extreme outlier (e.g., King Pukapuka) sharply inflates the standard deviation.
  1. Units of Measurement:
  • Retains the exact same unit of measurement as the original raw data (e.g., minutes, pounds, tree counts).

• Example: Radon Levels in Franklin County, Ohio (Textbook 2.10):

  • Background: Radon gas is a naturally occurring radioactive gas formed by the decay of uranium in soil that enters buildings through foundation cracks. It is the second leading cause of lung cancer in the United States (smoking is the first).
  • Dataset: Four measured radon levels from Franklin County, Ohio:     $3.8, \quad 1.9, \quad 12.1, \quad 14.4$
  • Practice Standard: Manual computation of standard deviation via step-by-step variance tables is tedious and error-prone. All standard deviation calculations should be executed using statistical calculators.

Calculator Guide: TI-30X IIs Procedures

• Pre-Computation Display Check:

  • Ensure STAT is NOT displayed at the bottom of the screen before beginning. A display showing DEG is normal.

• Step-by-Step 1-Variable Data Entry:

  1. Press 2nd then STAT (3rd row, 3rd button).
  2. Select 1-VARIABLE by ensuring it is underlined, then press ENTER. The word STAT will now appear at the bottom of the screen.
  3. Press DATA (3rd row, 3rd button).
  4. Input observations:
  • x1 = 3.8, press down arrow. FRQ = 1 (leave frequency at 1), press down arrow.
  • x2 = 1.9, press down arrow. FRQ = 1, press down arrow.
  • x3 = 12.1, press down arrow. FRQ = 1, press down arrow.
  • x4 = 14.4, press down arrow. FRQ = 1.

• Retrieving Summary Statistics:

  1. Press STAT VAR (3rd row, 4th button).
  2. Under n: Displays sample size ($n = 4$). Always check this value to confirm correct entry count.
  3. Scroll right to x-bar ($\bar{x}$): Displays sample mean:      $\bar{x} = 8.05$
  4. Scroll right to Sx ($s_x$): Displays sample standard deviation:      $s_x = 6.1266$      (rounded to four decimal places).

• Clearing Data (EXIT STAT Procedure):

  1. Press 2nd then EXIT STAT (3rd row, 4th button).
  2. Screen displays Y N with Y underlined. Press ENTER to confirm clearing.
  3. Verify that STAT has disappeared from the bottom display. Never clear data by typing over old values.

Statistical Problem-Solving and Application: Logging in Borneo (Textbook 2.13)

• The Four-Step Process:

  • State: Ask the research or statistical question.
  • Plan: Determine the analytical methods and statistical tools to be used.
  • Solve: Execute calculations, construct graphs, and generate summary statistics.
  • Conclude: Interpret findings within the real-world context of the problem.

• Borneo Study Details:

  • Conducted by Charles Cannon of Duke University and co-workers.
  • Subject: Impact of tropical rainforest clearing, burning, and logging on tree counts in Borneo.
  • Design: Compared $0.1\,\text{hectare}$ forest plots across three distinct groups:
  • Group 1: Never logged.
  • Group 2: Logged $1\,\text{year}$ earlier.
  • Group 3: Logged $8\,\text{years}$ earlier.

• Summary Results Across Groups:

  • Group 1 (Never Logged):$\bar{x} = 23.75\,\text{trees}, \quad s_x = 5.0655\,\text{trees}$
  • Group 2 (Logged 1 Year Ago):$\bar{x} = 14.0833\,\text{trees}, \quad s_x = s$
  • Group 3 (Logged 8 Years Ago - $n = 9$ plots):$\bar{x} = 15.7778\,\text{trees}, \quad s_x = 5.7615\,\text{trees}$

• Contextual Interpretation:

  • Standard deviations across all three groups are comparable ($s \approx 5$), validating comparative analysis across treatment conditions.
  • Unlogged virgin plot counts average roughly $24\,\text{trees}$.
  • Immediate post-logging (1 year) counts drop sharply to roughly $14\,\text{trees}$.
  • After 8 years, tree counts show negligible recovery (rising only to roughly $15.78\,\text{trees}$), demonstrating that tropical rainforest regeneration following logging is an extraordinarily slow process.

Questions and Discussion

• Student Question on Exam Format and Data Workflow:

  • Prompt: Will exam questions provide raw data requires constructing a stemplot prior to computing summary statistics?
  • Response: Yes. Exams present raw data where students construct stemplots to order data, compute the Five-Number Summary manually, and utilize calculators to find the mean and sample standard deviation.

• Student Question on Patriots Dataset Minimum:

  • Prompt: Was the minimum weight value in the Patriots dataset equal to $215\,\text{lbs}$?
  • Response: No. The exact minimum weight value in the dataset is strictly $250\,\text{lbs}$. (There is no $215\,\text{lbs}$ value in the dataset, though a $315\,\text{lbs}$ value is present).