TYPES OF SAMPLING DESIGN

Probability Sampling

One sampling unit has a known and equal probability of being selected compared with the other samples in the sampling frame

. Non-probability Sampling

One sampling unit does not have known and equal probability of being selected compared with the other samples in the sampling frame

. Probability sampling

The type of sampling where it is possible to determine which sampling units belong to which sampling frame and the probability each will be selected

. Simple Random Sampling (SRS), Systematic Sampling, Stratified Sampling, Cluster Sampling, Multistage Sampling

Examples of probability sampling methods

. Simple Random Sampling (SRS)

The simplest probability sampling method where each sampling unit in the sampling frame has equal chances of selection

. Fish bowl method, random number tables, computer-based (pseudo-)random number generators

Methods historically or recently used for Simple Random Sampling

. Systematic Sampling

A variation of Simple Random Sampling where only the starting point is randomized

. Systematic Sampling

The sampling method where samples after the starting point are selected using a calculated and fixed interval based on a specific formula

. Stratified Sampling and Cluster Sampling

Two probability sampling methods most people have trouble distinguishing

. Stratified Sampling and Cluster Sampling

Sampling methods that use specific strategies to segregate the sample population into different groups

. Groups are defined differently

The main distinction between Stratified Sampling and Cluster Sampling groups

. Stratified Sampling

Sampling method used when it makes sense to partition the population into groups (strata) based on a factor that may influence the variable being measured

. Strata

The name for the groups partitioned in Stratified Sampling

. Stratum

The name for an individual group in Stratified Sampling

. Partition the population into groups (strata)

The first step in Stratified Sampling

. Obtain a simple random sample of X sampling units from each group (stratum)

The second step in Stratified Sampling

. Collect data on each sampling unit that was randomly sampled from each group (stratum)

The third step in Stratified Sampling

. When a population is split into strata which are heterogeneous compared to other strata, but their sampling units are homogeneous

Conditions under which Stratified Sampling works best

. Strata 1 = all males, Strata 2 = all females; sampling units are the same in each strata, but strata 1 is only males and strata 2 only females

An example illustrating heterogeneous strata but homogeneous sampling units within strata for Stratified Sampling

. Stratified Sampling

Sampling method that generally produces more precise estimates of population percents than simple random sampling under certain conditions

. Cluster Sampling

Sampling method that is very different from Stratified Sampling

. Divide the population into X groups (clusters)

The first step in Cluster Sampling

. Obtain a simple random sample of X many clusters from all possible clusters

The second step in Cluster Sampling

. Obtain data on every sampling unit in each of the randomly selected clusters

The third step in Cluster Sampling

. Microcosms

What clusters should be, rather than subsections of the population, in Cluster Sampling

. Each cluster should be homogeneous compared to each other, but cluster sampling units are heterogeneous

Characteristics of clusters and sampling units in Cluster Sampling

. Cluster 1 = Pasig city, Cluster 2 = Taguig city; sampling units composed of varying ages, sex, income status etc., but distribution is about the same in Pasig vs Taguig

An example illustrating homogeneous clusters but heterogeneous sampling units within clusters for Cluster Sampling

. Unlike with the strata in Stratified Sampling

A key difference in the nature of the groups used in Cluster Sampling

. Undercoverage bias

A bias inherent with Cluster Sampling

. Adjustment made

What is usually done to the statistical analysis used with Cluster Sampling to account for undercoverage bias

. Example 1 (potentially affected by time zone)

An example where Stratified Sampling would be preferred over Cluster Sampling

. Stratified Sampling

Preferred method for Example 1 if the questions are affected by time zone

. If you use cluster sampling and the selected cluster differs significantly in time zone from a live event

Reason results might be unusually low for Example 1 using Cluster Sampling

. Example 2 (Sport)

An example where either Stratified or Cluster Sampling could be used depending on the question

. Cluster Sampling

Sampling method usable for Example 2 if the question concerns the effect of ANY sport, regardless of type

. Stratified Sampling

Sampling method preferred for Example 2 if the question concerns the effect of playing sports where the answer is influenced by the type of sport

. Example 3 (School)

An example where Cluster Sampling would probably be better than Stratified Sampling

. Each cluster appropriately represents the entire population and you have limited resources

Conditions under which Cluster Sampling is better for Example 3

. Students from one school more or less have the same characteristics compared to students from other schools

An assumption about schools in Example 3 where Cluster Sampling is preferred

. Save time and resources

An advantage of Cluster Sampling in Example 3 by selecting a sample of schools instead of all

. Convenience, Quota, Judgmental (or Purposive)

Three common non-probability sampling methods

. Convenience sampling versus Purposive sampling

Two non-probability sampling techniques often confused for each other

. Total Enumeration or Total Population Sampling

A special type of sampling often classified as non-probability

. Census

Another term that is the same as Total Enumeration or Total Population Sampling

. Include the entire population as your sampling frame

Requirement of Total Enumeration / Total Population Sampling

. You can often generalize your results with the entire population

Advantage of Total Enumeration / Total Population Sampling

. If you fail to recruit all the members of the population in question

Disadvantage of Total Enumeration / Total Population Sampling that causes results and analysis to suffer

. Populations with a very low population count that can be recruited easily

When Total Enumeration / Total Population Sampling is usually reserved

. e.g., people with a rare disease who are housed in a special facility for management

An example of a population where Total Enumeration is usually used

. Non-probability sampling

Often argued to be inferior to probability sampling

. Inherent selection bias and often non-generalizability of results

Reasons why non-probability sampling is argued to be inferior

. The author of the text

Person who is personally biased against non-probability sampling and would push for probability sampling

. When you are doing a qualitative study

A theoretical and practical use for non-probability sampling

. Investigate specific groups of people or phenomena

The need addressed by non-probability sampling in qualitative studies

. Purposively select your respondents

What you NEED to do in qualitative study designs using non-probability sampling

. To study these phenomena or characteristics in detail, rather than trying to generalize them to a broader population group

The objective in qualitative study designs using non-probability sampling

. When you have limited resources or time, or if there are extreme logistical or security issues

Practical reasons to opt for non-probability sampling methods

. Relatively inexpensive and often more feasible to do

Advantages of non-probability sampling methods under limited resources, time, or logistical/security issues

. Using the sampling method that will best answer your research objectives using your study design

The primary factor to consider when selecting your sampling method

. Practical concerns (logistic, safety, financial)

Other factors to consider when selecting your sampling method

. Include any limitations in your sampling strategy in your manuscript

Important step to properly inform readers

. Probability Sampling

Category of sampling where one unit has a known and equal probability of selection

. Non-probability Sampling

Category of sampling where one unit does not have a known and equal probability of selection

. Sampling frame

The set of sampling units from which a sample is drawn

. Simple Random Sampling

Probability sampling method where each unit in the frame has equal chances of selection

. Systematic Sampling

Probability sampling method where only the start is random, and subsequent units are selected by a fixed interval

. Calculated and fixed interval

Used to select samples after the start in Systematic Sampling

. Stratified Sampling

Probability sampling method that partitions the population into strata

. Strata

Groups in Stratified Sampling based on factors influencing the variable being measured

. Simple random sample of X sampling units

What is obtained from each stratum in Stratified Sampling

. Heterogeneous

How strata should be compared to other strata in Stratified Sampling

. Homogeneous

How sampling units within a stratum should be in Stratified Sampling for best results

. Cluster Sampling

Probability sampling method that divides the population into groups called clusters

. Simple random sample of X many clusters

What is obtained from all possible clusters in Cluster Sampling

. Every sampling unit

What data is obtained on in each of the randomly selected clusters

. Homogeneous

How clusters should be compared to each other in Cluster Sampling

. Heterogeneous

How sampling units within a cluster should be in Cluster Sampling

. Account for undercoverage bias

Why adjustments are often made to statistical analysis with Cluster Sampling

. Time zone

A factor that might influence the variable being measured in Example 1

. Watching a live sporting event on television

An example of a question affected by time zone in Example 1

. Question about the effect of ANY sport on grades

A question in Example 2 suitable for Cluster Sampling

. Question about whether playing sports increases chances of injury

A question in Example 2 requiring Stratified Sampling

. Type of sport each team is playing

Factor influencing the answer to the injury risk question in Example 2

. Appropriately represents the entire population

A condition for clusters in Example 3 that favors Cluster Sampling

. Limited resources

A practical constraint favoring Cluster Sampling in Example 3

. Schools

The clusters in Example 3

. Convenience Sampling

A common non-probability sampling method

. Quota Sampling

A common non-probability sampling method

. Judgmental (or Purposive) Sampling

A common non-probability sampling method

. Total Enumeration

Sampling technique requiring the entire population as the sampling frame

. Census

Synonym for Total Enumeration

. Generalize your results with the entire population

Advantage of Total Enumeration

. Fail to recruit all members

Disadvantage of Total Enumeration

. Qualitative study

Study type where non-probability sampling is often used

. Investigate specific groups or phenomena

Objective of qualitative studies suited for non-probability sampling

. Purposively select

Action taken when choosing respondents in qualitative studies

. Detail

How specific phenomena or characteristics are studied in qualitative designs

. Limited resources, time, logistical/security issues

Practical constraints favoring non-probability sampling

. Inexpensive and feasible

Characteristics of non-probability sampling methods under practical constraints

. Research objectives and study design

Primary drivers for selecting a sampling method