Unit 2: Sampling

1. Sampling: Meaning

In survey research, it is usually impossible to talk to everyone in a large group (the **"Population"**). For example, you cannot interview all 1.4 billion people in India.

Definition: Sampling is the scientific process of selecting a small, representative subset of a population (the **"Sample"**) to be studied. The findings from the sample are then used to make generalizations about the entire population.

• Population (or Universe): The entire group you want to study (e.g., "All voters in India").
• Sample: The small group you actually survey (e.g., "5,000 voters selected from across India").
• The Goal: To select a sample that is a "miniature version" of the population, so your findings are accurate and unbiased.

2. Types of Sampling

There are two main branches of sampling methods: **Probability Sampling** (which is scientific and random) and **Non-Probability Sampling** (which is based on convenience or judgment).

3. Probability (Random) Sampling

In this method, every member of the population has a known, non-zero chance of being selected. This is the "gold standard" for scientific surveys.

Types of Probability Sampling:

1. Simple Random Sampling (SRS):
   • The most basic method. Each person has an equal chance of being picked.
   • Method: Like pulling names out of a hat or using a random number generator.
   • Pro: Very simple. Con: Can be inefficient; you might miss small groups by chance.
2. Systematic Sampling:
   • Picking every "nth" person from a list.
   • Method: Get a list of the population (e.g., 10,000 people). You need a sample of 100. You divide 10,000/100 = 100. So, you pick a random starting point (e.g., #27) and then select every 100th person (#27, #127, #227...).
   • Pro: Easier than SRS. Con: Risk of bias if the list has a hidden pattern.
3. Stratified Sampling:
   • Used when the population has distinct sub-groups (**"strata"**) that you want to represent accurately.
   • Method: First, divide the population into groups (e.g., by gender, by state, by age). Then, do Simple Random Sampling *within* each group.
   • Example: To survey a college, you first divide by department (Arts, Science, Commerce) and then randomly select 50 students *from each* department. This ensures all departments are represented.
   • Pro: Highly representative. Con: Complex and requires prior knowledge of the population.
4. Cluster Sampling:
   • Used when the population is geographically widespread.
   • Method: Divide the population into "clusters" (e.g., districts, city blocks). Randomly select a few *clusters*. Then, survey *everyone* within those selected clusters.
   • Example: To survey a city, randomly pick 10 city blocks (clusters) and then interview every household in those 10 blocks.
   • Pro: Saves time and money. Con: Less precise than Stratified.

4. Non-Probability Sampling

In this method, selection is not random. It is based on the researcher's convenience, judgment, or a quota. It's cheaper and faster but cannot be scientifically generalized.

Types of Non-Probability Sampling:

1. Convenience Sampling:
   • Selecting whoever is easiest to reach.
   • Example: A TV reporter interviewing people at a mall.
   • Pro: Fast, cheap. Con: Highly biased, not representative.
2. Quota Sampling:
   • The non-probability version of stratified sampling.
   • Method: The researcher has a "quota" to fill (e.g., "I need 20 men and 30 women"). They then find these people using convenience methods.
   • Example: An interviewer stands on a street and stops people until they have interviewed 20 men and 30 women.
   • Pro: Ensures key groups are included. Con: Still biased, as the selection is not random.
3. Purposive (or Judgmental) Sampling:
   • The researcher uses their "judgment" to select people who are "experts" or typical of a group.
   • Example: If studying a village, you "purposively" select the village headman and the schoolteacher to get an expert view.
4. Snowball Sampling:
   • Used for hard-to-reach populations (e.g., homeless people, drug users).
   • Method: You find one or two people and then ask them to "refer" you to their friends, who then refer you to *their* friends. The sample "snowballs."

5. Sampling Error

Definition: Sampling error is the unavoidable statistical inaccuracy that results from surveying a sample instead of the entire population. It is the calculated "margin of error" in a poll.

• What it means: A poll might report "45% support for Party A, with a 3% margin of error." This means the *real* support in the population is likely between 42% (45-3) and 48% (45+3).
• Cause: It's the natural result of "chance." By pure luck, your random sample might have slightly more men, or slightly more old people, than the population.
• Key Fact: Sampling error can be *reduced* by increasing the **sample size**. A larger sample (e.g., 20,000) will have a smaller error than a small sample (e.g., 500).
• Note: This is different from "Non-Sampling Error" (e.g., bad questions, biased interviewers), which cannot be measured and can make a poll completely wrong.

6. Exam Corner: Key Distinctions