Sampling (6.1)

Main purpose of sampling:

Choose a representative subgroup of the population for testing/research

• used to draw accurate conclusions about the population

Sampling considerations:

• Research question

• Characteristics of wanted population

• Concept/idea you want to study

• Access to populations

Steps of Sampling plan

Define population > Define Sample > Determine Sample size and approach > Implementing sampling procedures

Sample definition:

Should accurately reflect the population.

• Accessible population should be considered

• Selection criteria

2 types of Selection criteria

Inclusion Criteria

Exclusion Criteria

Inclusion criteria:

characteristics participants MUST have

Exclusion criteria:

characteristics participants MUST NOT have

Sampling Error:

difference between data from sample and actual data existing in population

• systemic or random errors

Larger sampling error…

= less external validity

Sampling bias:

Can cause sample error

• conscious or unconscious

Is having lots of participants better?

Yes but that is not always achievable

2 Types of Errors:

Type I Error

Tupe II Error

Type I Error:

False-Positive

• Investigator rejects null hypothesis that is actually true in the population

• Mistakenly thinking there’s an effect

Type II Error:

False-Negative

• Investigator fails to reject a null hypothesis that is actually false in the population

• Mistakenly thinking there’s NO effect

Null Hypothesis (H₀)

There is NO relationship between variables

• Assumed to be true unless proven otherwise

Alternative Hypothesis (H₁)

There IS a relationship between variables

• Needs to be proven

4 ways to avoid Type I and Type II Errors:

• Sample size

• Statistical Power

• Effect Size

• Level of Significance

Power Analysis looks at:

Statistical Power, Effect size, Level of significance

Statistical Power

Likelihood of detecting a difference or association

• High power indicates large chance of effect

Effect Size (Cohen’s d, r, etc.)

Strength of difference or association

• Large effect size needs smaller sample

Level of Significance (p<.05)

There is a 5% chance that there is a mistake in determining effect

• There is 95% chance the effect was true

Pragmatic definition:

practical, and sensible based on real world conditions

Pragmatic considerations of Sample size and approach:

Larger sample

• Larger commitment to research time, effort, cost

Consider smallest to run statistical analysis through power analysis

Probability:

Each person has equal and independent chance of selection

• random sampling

• reduces potential bias

Non-Probability

Not feasible

• non-random sampling

• Goal: to attain greatest degree of representation

• Bias or error may exist

• Need to acknowledge limitations

4 Types of probability sampling methods:

• Simple Random

• Systematic

• Stratified Random

• Cluster

Simple random sampling

everyone has the same chance

• Probability

Systematic sampling:

fixed interval

ex: every 10th person

• Probability

Stratified random sampling:

divide people into smaller groups, then picking random samples from each group

ex: split group into male and female. selecting a certain amount from each

• Probability

Cluster sampling:

divide a population in naturally occurring groups

• use random method to pick a few people from each group

4 Types of Non-Probability samples:

• Convenience

• Quota

• Purposive

• Snowball/Network

Convenience sample:

Selecting people that are easy to access, close by, or readily available

• Non-Probability

Quota sample:

split the target population into specific categories (age, gender, income) and known traits

• gather data from available individuals who fit the criteria until each quota is full for each group

• Non-Probability

Purposive sample:

Intentionally choosing participants based on traits, knowledge, or criteria

• Intentional choice, with predefined qualifications

• Non-Probability

Snowball/Network Sampling:

Existing study participants help researchers find and recruit future subjects from their own social networks

• Non-Probability