Sampling methods refer to the techniques used to select a subset of individuals or items from a larger population for research or data collection. There are various sampling methods, broadly categorized into probability sampling and non-probability sampling techniques. Let's explore them in more detail:
Probability Sampling Techniques: Simple Random Sampling: In this technique, each member of the population has an equal chance of being selected. Random selection can be achieved through methods like lottery or random number generation. Systematic Sampling: This method involves selecting every nth individual from the population after randomly selecting a starting point. For example, if the population size is 1,000 and the desired sample size is 100, every 10th individual could be selected. Stratified Sampling: The population is divided into homogeneous subgroups or strata, and individuals are randomly selected from each stratum in proportion to their representation in the population. This ensures representation from each subgroup. Cluster Sampling: The population is divided into clusters or groups, and a random sample of clusters is selected. Then, all individuals within the selected clusters are included in the sample. Multi-stage Sampling: This method combines two or more sampling techniques. For instance, it could involve using stratified sampling to select clusters and then systematic sampling within those clusters.
Non-Probability Sampling Techniques: Convenience Sampling: Participants are selected based on their accessibility or availability to the researcher. This method is convenient but may introduce bias since it relies on individuals who are easy to reach. Purposive Sampling: Participants are deliberately selected based on specific criteria defined by the researcher. This technique is useful when seeking individuals with certain characteristics or expertise. Snowball Sampling: Initially, a small group of participants is selected, and then they help recruit additional participants who meet the research criteria. This method is often used when studying hard-to-reach or hidden populations. Quota Sampling: Similar to stratified sampling, the population is divided into subgroups. However, instead of random selection, the researcher sets quotas for each subgroup to ensure representation. Volunteer Sampling: Participants self-select by responding to an invitation or call for volunteers. This method can introduce bias as it relies on individuals who are motivated to participate.
Probability sampling techniques aim to provide each member of the population with a known chance of being included in the sample, enabling statistical inference. Non-probability sampling techniques, on the other hand, do not guarantee a known probability of selection and are more commonly used when probability sampling is not feasible or practical.
It's important to note that the choice of sampling method depends on the research objectives, available resources, population characteristics, and the level of precision and generalizability required for the study. Each method has its strengths and limitations, and researchers must consider these factors when selecting the most appropriate sampling technique.
"Major differences between basis of sampling methods." -
Well, regarding inference, there are two different "approaches": (1) the probability-of-selection-based approach, and (2) the model-based/prediction approach. The former requires probability sampling, as that is the basis for inference. Each member of the sample would 'represent' a given portion of the population such that an infinite number of sample draws would, on average, exactly infer to the population. However, the latter only requires that the population or subpopulation/stratum to which a model is applied is appropriately described by that model. Any model failure should result in acceptably small bias. This can be applied to probability or nonprobability sampling. A balanced sample (say, perhaps balanced on the mean of the predictor or each of the predictors) would likely serve better than a random sample to adjust a model to the population. This works well when you are dealing with repeated samples and an occasional census which provides a very good predictor for each item on the sample, i.e. the same item on a previous census is the predictor for each variable of interest.
With more challenging nonprobability samples (regarding inference), methods have been developed to arrive at a more complex model, or to arrive at pseudo-weights for quasi-random sampling, both based on perhaps numerous covariates. For example, see "Comparing Alternatives for Estimation from Nonprobability Samples," by Richard Valliant, December 2019, Journal of Survey Statistics and Methodology 8(2),
DOI: 10.1093/jssam/smz003
https://doi.org/10.1093/jssam/smz003
Some methods are used in combination, such as calibration where a survey weight, the inverse probability-of-selection, is modified by adjusting to known totals of auxiliary variable data, and a regression weight (based on heteroscedasticity) may figure in as well.