Types of Sample Designs
Sampling designs can be broadly classified based on their representation basis into two main types: probability sampling and non-probability sampling.
Probability Sampling
Probability sampling is based on the concept of random selection. In this design, every item or element in the universe has an equal and independent chance of inclusion in the sample.
This approach allows researchers to measure estimation errors and the significance of results.
Probability sampling ensures the "law of Statistical Regularity," meaning that a randomly chosen sample will likely have the same composition and characteristics as the universe, making it the best technique for selecting a representative sample.
In quantitative research, randomisation is specifically used to avoid bias in sample selection and ensure representativeness.
Example :
- if we have to select a sample of 300 items from a universe of 15,000 items, then we can put the names or numbers of all the 15,000 items on slips of paper and conduct a lottery.
Non-probability Sampling
Non-probability sampling is 'non-random' sampling. This procedure does not provide a basis for estimating the probability that each item in the population has of being included in the sample.
Also known as deliberate, purposive, or judgement sampling, items are chosen deliberately by the researcher. The researcher's judgement is supreme, and they aim to select units that are typical or representative of the whole.
A significant drawback is that personal bias can easily enter the selection process, and sampling error cannot be estimated.
It is used when the number of elements in a population is unknown or cannot be individually identified, with selection depending on other considerations.
Non-probability designs are primarily used in qualitative research
Quota sampling : Interviewers choose participants based on quotas for various characteristics.
Complex Random Sampling Designs
Complex random sampling designs, also referred to as 'mixed sampling designs,' often combine elements of both probability and non-probability sampling procedures to select a sample. These designs are particularly useful in large-scale inquiries or when a simple random sample is impractical or too costly.
Here are some commonly used complex random sampling designs:
- Systematic sampling
- Stratified sampling
- Cluster sampling
- Area sampling
- Multi-stage sampling
- Sampling with probability proportional to size
- Sequential sampling
1. Systematic sampling
Only the first unit is selected randomly, and the remaining units of the sample are selected at fixed intervals.
This design is classified as a 'mixed' sampling design because it incorporates characteristics of both random and non-random approaches.
The process involves dividing the sampling frame into a number of segments called intervals. Then, a starting element is randomly selected from the first interval using a Simple Random Sampling (SRS) technique.
Subsequent elements from other intervals are selected based on the order of the element chosen in the first interval (e.g., if the fifth element was chosen initially, then the fifth element of every subsequent interval is chosen).
Selection at regular intervals; e.g., every 10th household.
2. Stratified sampling
This technique is applied when the population is not homogeneous and aims to obtain a more representative sample and achieve greater accuracy.
The population is first divided into several non-overlapping subgroups or 'strata' based on a characteristic that makes each subgroup more homogeneous.
After stratification, samples are selected from each stratum, typically using the SRS technique.
Population divided into strata (e.g., urban/rural), and random samples drawn from each.
3. Cluster sampling
When it is difficult or expensive to identify every element in a large population. The population is divided into groups called 'clusters' (based on visible or easily identifiable characteristics like geographical proximity or a common characteristic correlated with the study's main variable).
Instead of selecting individual elements, the researcher randomly selects a number of these clusters. The ultimate sample then consists of all units (or a sample of units) within the selected clusters.
Groups or clusters are selected instead of individuals (e.g., classrooms, neighborhoods).
- Departmental store wishes to sample its credit card holders. It has issued its cards to 15,000 customers. The sample size is say 450. Among 15,000 customers, 100 clusters are formed. Each cluster consists of 150 card holders. Three clusters are selected for the sample randomly
4. Area sampling
Area sampling is a specific type of cluster sampling where the clusters are geographical subdivisions. The total geographical area of interest is divided into smaller, non-overlapping areas (geographical clusters), from which a number are randomly selected. All units in these small areas are included in the sample.
5. Multi-stage sampling
This is a further development of cluster sampling, used for large inquiries extending over a considerable geographical area, such as an entire country. It involves sampling at different levels or stages.
First stage may be to select large primary sampling units such as states, then districts, then towns and finally certain families within towns. If random sampling is applied at all stages, it is called multi-stage random sampling.
Sampling in successive stages (e.g., states → districts → towns → households).
6. Sampling with probability proportional to size
This technique is used in cluster sampling when the clusters do not have approximately the same number of elements. The selection process ensures that the probability of each cluster being included in the sample is proportional to its size (i.e., the number of elements it contains).
7. Sequential sampling
This is a more complex sample design where the ultimate size of the sample is not fixed in advance. Instead, it is determined by mathematical decision rules based on information gathered as the survey progresses.
In sequential sampling, data collection can continue, with samples taken one after another, until a decision threshold is met or a desired saturation point is reached.
Selection of Appropriate Method for Data Collection
A researcher must judiciously select the data collection method for their study, keeping the following factors in view:
Nature, Scope, and Object of Enquiry: This is the most important factor. The selected method should suit the type of inquiry. This factor also helps decide whether to use primary or secondary data.
Availability of Funds: The research budget determines to a large extent the method to be used. Limited funds may require selecting a cheaper method that might not be as effective.
Time Factor: The time available also influences the choice of method, as some methods are more time-consuming than others.
Precision Required: The level of precision needed is another important factor.
No single method is superior in all situations. For example, a telephone interview might be appropriate if funds and time are limited and only a few items of data are needed. If funds permit and more information is desired, a personal interview may be better. Secondary data can be used if it is found to be reliable, adequate, and appropriate. Projective techniques are valuable in market research or psychological surveys, especially when a respondent knows the reason for something but is unwilling to admit it, or is unaware of their own underlying psychological attitudes.
Experiments vs. Surveys
An experiment is an investigation where a factor or variable is isolated and its effects are measured. The investigator intentionally conducts the experiment and measures its effects.
A survey is a method of securing information about a phenomenon from all or a selected number of respondents within the concerned universe. The investigator examines phenomena that exist independently of their actions.
Difference between Surveys and Experiments
Research Type:
Surveys are primarily employed in descriptive research studies, which focus on describing the current state of affairs, conditions that exist or existed.
Experiments are an integral part of experimental research studies (also known as hypothesis-testing research), which aim to clarify cause-and-effect relationships.
Sample Size
Surveys typically require larger samples because the percentage of responses can be low(as low as 20–30%).
Experiments generally need smaller samples.
Focus:
Surveys are concerned with describing, recording, analyzing, and interpreting existing or past conditions.
Experimental research provides a systematic method for answering "what will happen if this is done" when certain variables are controlled or manipulated.
Field of Application:
Surveys are generally more appropriate for the social and behavioral sciences, as many types of human behavior that researchers are interested in cannot be realistically arranged or controlled in an experimental setting.
Experiments are an essential feature of the physical and natural sciences, where strict control over conditions can often be enforced, typically in a laboratory environment.
Research Environment:
Surveys are examples of field research, conducted in natural settings.
Experiments generally constitute laboratory research, although they can also be carried out in natural environments depending on the study design.
Hypothesis Testing
Surveys are used for hypothesis formulation and testing by analyzing relationships between variables that are observed but not manipulated.
Experimentation directly provides a method for testing a proposed hypothesis. After defining a problem, experimenters propose a hypothesis, then test it and either confirm or disconfirm it.