1. Introduction to Sampling and Data
Welcome to the gateway of statistics! Before we can calculate means, model real-life trends, or find probabilities, we need to gather data. The quality of your conclusions depends entirely on the quality of your raw data. If your data collection method is flawed, even the most sophisticated mathematics will lead to invalid conclusions.
In this chapter, you will master the foundational concepts of SL 4.1: distinguishing between populations and samples, classifying different data types, evaluating sampling techniques, spotting sampling bias, and identifying mathematical outliers.
2. Population vs. Sample
Imagine you are cooking a massive pot of soup. You do not need to drink the entire pot to know if it needs salt; you simply stir it well and taste a single spoonful. This analogy captures the fundamental relationship between a population and a sample.
Key Definitions
Population: The entire group of individuals, objects, or items that you wish to investigate and draw conclusions about.
Example: All \(1\,200\) students enrolled at an international school.
Sample: A selected subset of the population from which data is actually collected.
Example: A group of \(60\) students chosen across different grades to complete a survey.
Random Sample: A sample chosen in such a way that every individual item or person in the population has an equal chance of being selected, and every combination of individuals has an equal chance of being chosen together.
Why Not Always Survey the Entire Population?
A complete survey of an entire population is called a census. While a census provides complete information, statisticians often rely on a sample because a census is:
• Expensive and time-consuming: Gathering data from thousands or millions of individuals requires massive resources.
• Impractical or impossible: Some populations are infinite, constantly changing, or geographically unreachable.
• Destructive: If quality testing involves testing battery life until the battery dies, testing the entire population would destroy every product!
Key Takeaway: A sample should be a miniature reflection of the population—large enough and chosen carefully to represent the population without introducing systematic distortion.
3. Data Types: Discrete vs. Continuous
Data gathered in statistical investigations can be categorical (qualitative) or numerical (quantitative). In IB Mathematics: Analysis and Approaches, our focus is primarily on numerical data, which is divided into two vital categories: discrete and continuous.
Discrete Data
Discrete data can only take specific, distinct values. There are distinct gaps between possible values, and the data is typically obtained by counting.
• Examples: The number of siblings a student has (\(0, 1, 2, 3, \dots\)), the score on a rolled die (\(1, 2, 3, 4, 5, 6\)), or shoe sizes (\(7, 7.5, 8, 8.5\)).
• Notice: Even though shoe sizes can have fractions like \(7.5\), shoe sizes are still discrete because you cannot have a shoe size of \(7.2341\). There are gaps between permitted values!
Continuous Data
Continuous data can take any numerical value within a given interval or range. Continuous data is obtained by measuring, and its precision is limited only by the measuring instrument used.
• Examples: The height of an athlete (\(175.4\text{ cm}\)), the time taken to run a \(100\text{ m}\) sprint (\(11.38\text{ s}\)), or the mass of an apple (\(142.85\text{ g}\)).
• Notice: Between any two values of continuous data, there is an infinite number of possible intermediate values.
Memory Aid: Count vs. Measure
• If you count it, it is discrete (e.g., number of cars in a car park).
• If you measure it with a ruler, scale, or stopwatch, it is continuous (e.g., length of journey, temperature).
Key Takeaway: Discrete data consists of separate, distinct points. Continuous data forms an unbroken scale of measurement.
4. Sampling Techniques and Their Effectiveness
To avoid bias and ensure that a sample accurately reflects the population, statisticians use specific sampling strategies. You must understand how each method works, when to use it, and its relative advantages and limitations.
1. Simple Random Sampling (SRS)
How it works: Every member of the population is assigned a unique number from \(1\) to \(N\). A random number generator or lottery system is then used to select \(n\) unique numbers without replacement.
• Strengths: Completely unbiased; free from researcher influence; every individual has an equal probability of selection.
• Limitations: Requires a full, up-to-date sampling frame (a complete list of every member of the population), which can be costly or impossible to obtain. It can also produce an unrepresentative sample by pure chance if the sample size is small.
2. Systematic Sampling
How it works: From a numbered list of the population of size \(N\), elements are chosen at regular intervals. Calculate the sampling interval \(k = \frac{N}{n}\). Choose a random starting point between \(1\) and \(k\), and then select every \(k\)th individual thereafter.
• Example: In a school of \(N = 800\) students, you want a sample of \(n = 40\). The interval is \(k = \frac{800}{40} = 20\). Select a random number between \(1\) and \(20\), say \(14\). Your sample includes students numbered \(14, 34, 54, 74, \dots, 794\).
• Strengths: Simple, quick, and spreads the sample evenly across the list.
• Limitations: If the list has an underlying periodic pattern or cycle that coincides with \(k\), the sample can become severely biased.
3. Stratified Sampling
How it works: The population contains distinct, non-overlapping subgroups called strata (such as age brackets, genders, or year levels) that may have different characteristics. Members are randomly selected from each stratum in direct proportion to the stratum's size relative to the entire population.
Formula for Stratum Sample Size:
\(\text{Number selected from stratum} = \frac{\text{Size of stratum}}{\text{Total population size}} \times \text{Total sample size}\)
• Example: A school has \(600\) boys and \(400\) girls (total \(1\,000\) students). For a sample of \(50\):
Number of boys \(= \frac{600}{1\,000} \times 50 = 30\)
Number of girls \(= \frac{400}{1\,000} \times 50 = 20\)
Within each group, the \(30\) boys and \(20\) girls are chosen using simple random sampling.
• Strengths: Guarantees proportional representation of key subgroups; avoids under-representing minority groups; gives high precision.
• Limitations: You must know the exact composition of the population beforehand, and categorizing individuals can be complex.
4. Quota Sampling
How it works: Similar to stratified sampling, the population is divided into subgroups. However, within each subgroup, the researcher selects participants non-randomly (often opportunistically) until the required quota for each group is filled.
• Strengths: Fast, practical, inexpensive, and does not require a complete list of every individual in the population.
• Limitations: Non-random. The interviewer chooses who to interview, which introduces interviewer bias and invalidates statistical probability theory.
5. Convenience Sampling
How it works: The researcher simply selects individuals who are easiest to reach or readily available at a particular time and place.
• Example: Standing outside the school cafeteria at 12:00 PM and interviewing the first \(30\) students who walk by.
• Strengths: Extremely cheap, fast, and easy to conduct.
• Limitations: Highly prone to bias; the sample is rarely representative of the wider population.
Key Takeaway: Random methods (Simple Random, Systematic, Stratified) eliminate researcher selection bias and allow valid statistical generalizations. Non-random methods (Quota, Convenience) are easier and faster, but carry a high risk of bias.
5. Reliability of Sources and Bias in Sampling
In IB exam questions, you may be asked to evaluate a sampling method and identify potential sources of bias or comment on the reliability of sources.
What is Sampling Bias?
Bias is a systematic distortion in the data collection process that causes the sample to favor certain outcomes or subgroups over others, leading to false conclusions.
Common Forms of Bias
• Selection Bias (Sampling Frame Error): The sampling frame (list used) does not match the actual population. For example, conducting an online survey about internet access excludes people without internet.
• Non-Response Bias: Individuals selected for the sample refuse to respond or cannot be contacted. Those who do not respond may hold systematically different views from those who do.
• Voluntary Response Bias: Occurs when participants volunteer themselves (e.g., call-in radio polls, open website polls). Typically, only people with strong or extreme opinions take part.
• Response / Questionnaire Bias: The wording of the question influences the respondent's answer (leading questions), or the respondent gives socially acceptable answers rather than the truth.
Evaluating Reliability of Sources
When critiquing data or evaluating whether a source is trustworthy, consider:
1. Who funded or collected the data? Is there an inherent commercial or political conflict of interest?
2. What was the sample size? Very small samples suffer from large random error and high variability.
3. How was the sample selected? Was it a proper random method or a biased convenience sample?
4. Is the full methodology transparent? Are questions, dates, and response rates clearly reported?
Key Takeaway: A large sample does not fix a biased method. A biased sample of \(100\,000\) people is still untrustworthy!
6. Outliers: Concept and Mathematical Detection
An outlier is an extreme data value that lies unusually far away from the main body of the data set. Outliers can arise from measurement errors, data entry mistakes, or genuine, rare natural variation.
The \(1.5 \times \text{IQR}\) Rule
In IB DP Mathematics: Analysis and Approaches, you must know how to identify outliers using the formal interquartile range (\(\text{IQR}\)) rule. The quartiles \(Q_1\) (lower quartile) and \(Q_3\) (upper quartile), and the measure \(\text{IQR} = Q_3 - Q_1\), are studied in detail when presenting and summarizing data.
Any data value \(x\) is classified as an outlier if it satisfies either of the following conditions:
Lower Outlier Boundary:
\(x < Q_1 - 1.5 \times \text{IQR}\)
Upper Outlier Boundary:
\(x > Q_3 + 1.5 \times \text{IQR}\)
Step-by-Step Procedure to Test for Outliers
• Step 1: Identify the lower quartile (\(Q_1\)) and upper quartile (\(Q_3\)).
• Step 2: Calculate the interquartile range: \(\text{IQR} = Q_3 - Q_1\).
• Step 3: Multiply the \(\text{IQR}\) by \(1.5\).
• Step 4: Calculate the lower boundary: \(Q_1 - 1.5(\text{IQR})\).
• Step 5: Calculate the upper boundary: \(Q_3 + 1.5(\text{IQR})\).
• Step 6: Compare the suspected data value against these boundaries. If it is strictly less than the lower boundary or strictly greater than the upper boundary, it is an outlier.
What Should You Do With an Outlier?
Never automatically erase an outlier! You should investigate its origin:
• If it is an obvious transcription error or equipment fault (e.g., a human height entered as \(185\text{ m}\) instead of \(1.85\text{ m}\)), it should be corrected or removed.
• If it is a genuine, verified measurement (e.g., an exceptionally gifted athlete running an unusually fast sprint), it must be retained because it represents true natural variation in the population.
Key Takeaway: An outlier is mathematically defined by falling outside the interval \([Q_1 - 1.5(\text{IQR}),\, Q_3 + 1.5(\text{IQR})]\). Always check both boundaries when evaluating suspected values.
7. Worked Examples
Example 1: Sampling Calculations (Paper 1 Style)
A university has \(4\,500\) undergraduate students and \(1\,500\) postgraduate students. A researcher wants to take a stratified sample of \(120\) students to investigate study habits.
(a) Calculate the number of undergraduate students that should be included in the sample.
(b) Briefly describe how the researcher can select the required undergraduate students using simple random sampling.
(c) State one reason why stratified sampling is preferable to convenience sampling in this context.
Solution:
(a) Total population \(= 4\,500 + 1\,500 = 6\,000\).
\(\text{Number of undergraduates} = \frac{4\,500}{6\,000} \times 120 = \frac{3}{4} \times 120 = 90\).
Answer: \(90\) undergraduate students.
(b) Assign each of the \(4\,500\) undergraduate students a unique integer from \(1\) to \(4\,500\). Use a random number generator to generate \(90\) unique random integers in this range, ignoring any repeats. Select the students corresponding to those \(90\) numbers.
(c) Convenience sampling would only select students who are readily accessible (such as those in one specific lecture theatre or the campus cafe at a particular hour), which introduces bias and may exclude certain fields of study. Stratified sampling guarantees that both undergraduate and postgraduate cohorts are represented in exact proportion to their population sizes, eliminating researcher bias.
Example 2: Identifying Outliers Mathematically (Paper 1 / Paper 2 Style)
A set of test scores has a lower quartile of \(Q_1 = 48\) and an upper quartile of \(Q_3 = 72\).
Two students achieve scores of \(10\) and \(105\) respectively.
Determine mathematically whether either of these two scores is an outlier.
Solution:
Step 1: Calculate the \(\text{IQR}\)
\(\text{IQR} = Q_3 - Q_1 = 72 - 48 = 24\).
Step 2: Calculate \(1.5 \times \text{IQR}\)
\(1.5 \times 24 = 36\).
Step 3: Determine the lower and upper boundaries
\(\text{Lower boundary} = Q_1 - 1.5(\text{IQR}) = 48 - 36 = 12\).
\(\text{Upper boundary} = Q_3 + 1.5(\text{IQR}) = 72 + 36 = 108\).
Step 4: Test each value
• For score \(10\): Since \(10 < 12\), the score lies below the lower boundary. Therefore, \(10\) is an outlier.
• For score \(105\): Since \(105 \le 108\), the score does not exceed the upper boundary. Therefore, \(105\) is not an outlier.
8. Common Mistakes and Exam Tips
• Mistake: Confusing discrete data with integers only. Shoe sizes (\(8.5, 9, 9.5\)) are discrete even though they contain decimals. The defining feature is the presence of distinct steps or gaps, not whether the numbers are whole integers.
• Mistake: Misremembering the outlier multiplier. The standard IB definition uses \(1.5 \times \text{IQR}\), not \(2 \times \text{IQR}\) or \(1 \times \text{IQR}\).
• Mistake: Vague justifications in sampling questions. When asked why a sample is biased, do not simply write "it is not fair." State specifically who is excluded or over-represented (e.g., "Surveying people at a train station at 8:00 AM over-represents daily commuters and excludes retirees, students, and home workers").
• Exam Tip: Always show the numerical boundaries. When testing for an outlier, always write down the explicit values of \(Q_1 - 1.5(\text{IQR})\) and \(Q_3 + 1.5(\text{IQR})\) before stating your conclusion. IB markschemes award method marks for calculating these threshold boundaries!
9. Chapter Summary Quick Review
• Population: The entire group of interest.
• Sample: A representative subset of the population.
• Discrete Data: Counted values with distinct gaps.
• Continuous Data: Measured values along an unbroken numerical continuum.
• Simple Random Sampling: Every member and combination has an equal chance of selection.
• Systematic Sampling: Select every \(k\)th item after a random starting point in the first interval.
• Stratified Sampling: Randomly sampled within subgroups in exact proportion to population size.
• Quota Sampling: Stratified representation filled through non-random selection.
• Convenience Sampling: Selecting the most accessible individuals; fast but highly biased.
• Outlier Boundaries: Values strictly less than \(Q_1 - 1.5(\text{IQR})\) or strictly greater than \(Q_3 + 1.5(\text{IQR})\).