Sampling Techniques: Surveys and Population Characteristics

TL;DR

You'll learn how to pick a small group (sample) to understand a larger group (population) without needing to study everyone. Choosing the right sampling method is crucial for getting reliable results and avoiding misleading conclusions. This topic helps you make sure your survey findings accurately reflect the population you're interested in.

1. The Mental Model

Imagine you want to know how sweet a giant bowl of punch is. You don't drink the whole bowl; you take a small sip. That sip is your sample, and the whole bowl is your population. The goal is for that sip to tell you a lot about the whole punch.

2. The Core Material

When you're doing a survey, you're usually trying to understand something about a population – that's the entire group of individuals or items you're interested in. For example, all adults in your city, or all medical records in a hospital. However, studying an entire population is often impossible, too expensive, or too time-consuming. That's why you use a sample – a smaller, manageable subgroup drawn from the population.

The key challenge is to make sure your sample is representative of the population. If it's not, your findings might be biased and not accurately reflect the larger group.

Types of Sampling

Photo by Gustavo Fring on Pexels

There are two main categories of sampling:

a) Probability Sampling

This is where every member of the population has a known, non-zero chance of being selected for the sample. This is generally preferred because it allows you to make statistically sound inferences about the population.

• Simple Random Sampling (SRS): Every possible sample of a given size has an equal chance of being selected. Imagine putting all names in a hat and drawing them out.

  • When to use: Small, homogeneous populations; when you have a complete list of the population (a "sampling frame").
  • Pros: Easy to understand, unbiased estimates.
  • Cons: Can be impractical for large populations, might not be representative if the population has distinct subgroups.

• Systematic Sampling: You select a random starting point and then pick every kth element from the population list. For example, if you want 100 people from a list of 1000, you'd pick every 10th person after a random start.

  • When to use: When you have a complete, ordered list and SRS is too cumbersome.
  • Pros: Simpler than SRS, good coverage of the population.
  • Cons: If there's a hidden pattern in the list that aligns with your kth interval, it can introduce bias.

• Stratified Sampling: You divide the population into distinct subgroups (strata) based on some shared characteristic (e.g., age groups, gender, disease severity). Then, you perform SRS within each stratum.

  • When to use: When your population is heterogeneous and you want to ensure representation from all important subgroups.
  • Pros: Ensures representation of key subgroups, can lead to more precise estimates than SRS.
  • Cons: Requires prior knowledge to define strata, more complex to implement.

• Cluster Sampling: You divide the population into clusters (e.g., geographical areas, hospitals). You then randomly select some clusters and include all individuals from those selected clusters in your sample.

  • When to use: When a complete list of individuals is unavailable, but clusters are easily identifiable (e.g., surveying all patients in selected clinics).
  • Pros: Cost-effective, especially for geographically dispersed populations.
  • Cons: Can be less precise than SRS or stratified sampling if clusters aren't very diverse internally.

b) Non-Probability Sampling

This is where selection is not random. It's often used when probability sampling isn't feasible or for exploratory research, but it makes it harder to generalize findings to the broader population.

• Convenience Sampling: You select individuals who are easiest to reach. For instance, surveying people walking by in a particular location.

  • Pros: Very easy and inexpensive.
  • Cons: High risk of bias, findings rarely generalize.

• Quota Sampling: Similar to stratified sampling, but instead of random selection within strata, you non-randomly select individuals until you meet a predefined quota for each subgroup (e.g., survey 50 men and 50 women).

  • Pros: Attempts to ensure representation of subgroups.
  • Cons: Still prone to selection bias within quotas.

• Purposive (Judgmental) Sampling: You select participants based on your expertise and judgment about who would be most useful or representative for your research question.

  • Pros: Useful for very specific populations or qualitative research.
  • Cons: Highly susceptible to researcher bias.

Here's a diagram to help visualize the different approaches:

graph TD
    A["Sampling Techniques"] --> B["Probability Sampling"]
    A["Sampling Techniques"] --> C["Non-Probability Sampling"]

    B --> D["Simple Random Sampling (SRS)"]
    B --> E["Systematic Sampling"]
    B --> F["Stratified Sampling"]
    B --> G["Cluster Sampling"]

    C --> H["Convenience Sampling"]
    C --> I["Quota Sampling"]
    C --> J["Purposive Sampling"]

Key Terms You'll Encounter

Photo by Ann H on Pexels

• Population Parameter: A characteristic of the entire population (e.g., the average height of all adults in a country). This is usually what you want to estimate.
• Sample Statistic: A characteristic calculated from your sample (e.g., the average height of people in your sample). You use this to estimate the population parameter.
• Sampling Bias: Occurs when your sample isn't truly representative of the population, leading to systematic errors in your estimates.
• Sampling Error: The natural variability that occurs when you take a sample instead of measuring the entire population. Even with perfect random sampling, your sample statistic won't exactly match the population parameter. This error can be quantified.

3. Worked Example

Let's say you're a hospital administrator and you want to estimate the average patient satisfaction score for all patients admitted last year (approximately 5000 patients). You can't survey every single patient.

Goal: Estimate average patient satisfaction (on a scale of 1-10) with a sample of 200 patients.

Method: Stratified Random Sampling.

1. Define Strata: You realize patient satisfaction might differ significantly between different departments (e.g., emergency, surgery, maternity). You decide to stratify by department.

   • Emergency: 1500 patients
   • Surgery: 2000 patients
   • Maternity: 1000 patients
   • Other: 500 patients
   • Total: 5000 patients

2. Calculate Proportional Allocation: To ensure representation, you'll sample patients from each department proportionally to their size in the population.

   • Emergency: (1500 / 5000) * 200 = 60 patients
   • Surgery: (2000 / 5000) * 200 = 80 patients
   • Maternity: (1000 / 5000) * 200 = 40 patients
   • Other: (500 / 5000) * 200 = 20 patients
   • Total Sample: 60 + 80 + 40 + 20 = 200 patients

3. Perform Simple Random Sampling within each Stratum: You get a list of all patients from each department from last year. For each department, you use a random number generator to select the required number of patients.

   • From the Emergency department list, randomly pick 60 patients.
   • From the Surgery department list, randomly pick 80 patients.
   • And so on for Maternity and Other.

4. Collect Data and Calculate Sample Statistic: You then survey these 200 selected patients and calculate their average satisfaction score. Let's say your sample's average satisfaction score is 8.2. You would then report that the estimated average patient satisfaction for the hospital is 8.2, and because you used probability sampling, you can also calculate a margin of error around this estimate.

4. Key Takeaways

• Sampling lets you learn about a large group (population) by studying a small part (sample).
• Probability sampling (like SRS, stratified, cluster) allows you to make reliable generalizations because every member has a known chance of being picked.
• Non-probability sampling (like convenience, quota) is easier but usually leads to less reliable, biased results.
• A representative sample is crucial; otherwise, your findings won't accurately reflect the population.
• Always aim for probability sampling when you want to make statistical inferences about a population.
• Understanding the difference between population parameters and sample statistics is fundamental.

Common Mistakes to Avoid:

• Convenience sampling for formal studies: Don't use "whoever is available" if you need to generalize your findings.
• Ignoring sampling bias: Always think about who might be over- or under-represented in your sample.
• Confusing population and sample statistics: Remember your sample mean is an estimate of the true population mean.
• Using too small a sample: While saving time, a very small sample size increases your sampling error and makes your results less precise.

5. Now Try It

Imagine you want to study the prevalence of smartphone use among college students at your university (approx. 20,000 students). You have access to the university's student directory, which lists all students, their major, and their year of study (freshman, sophomore, etc.). Design a sampling plan to select 500 students.

What to do:
1. State which probability sampling technique you would use and why.
2. Briefly describe the steps you would take to implement this plan.
3. Explain one potential advantage and one potential disadvantage of your chosen method for this scenario.

What success looks like: You've clearly identified a suitable probability sampling method, outlined logical steps for its implementation using the given information, and articulated its pros and cons in this specific context.