What are the disadvantages of cluster sampling

Sampling

Upper chapter: II. Planning and preparation of the survey

Sampling

In social research, due to time and cost reasons, it is in most cases not feasible to integrate the entire population into a planned survey (full census, census). Therefore, samples are used (partial survey). Samples represent a subset of all study units, which represent the properties of the population that are relevant for the research goal as precisely as possible.

Basically, the drawing of a sample is intended to be representative. This means that conclusions are drawn from the known characteristic values โ€‹โ€‹of the sample to unknown characteristic values โ€‹โ€‹(parameters) of the population.

Compared to a full census, sampling has the disadvantage that the selection of the sample can lead to a selection bias. In order to counteract such systematic distortions, careful sampling is of great importance.

A population is defined as the universe of all units. This population must be defined in connection with the research question. From these units, which can represent countries, regions or other associations of groups, a sample is selected in the context of empirical studies. Units of the population to which this selection (sample) relates and which have any chance of being included in the sample are referred to as survey units. If, for example, an empirical study is to be carried out on employee motivation in Swiss SMEs, all Swiss SMEs represent the population, and 5 selected SMEs per canton represent the survey units.

Type of sampling

As part of the sampling process, it must be decided which selection process will be used to define the subset of the population to be examined. A sampling procedure can in principle be specified in Art

  • Probability sampling,
  • conscious selection (non-probability sampling) or
  • arbitrary selection (non-probability sampling)

respectively.

Probability / random selection

The representativeness of the sample is given by the probability selection of samples, in contrast to the arbitrary selection and conscious selection. This means that the probability selection allows a generalization from the sample to the population. Such conclusions are made using statistical tests.

Probability selection can be done in three different ways: simple random sample, stratified sample, or cluster sample.

Simple random sample

The most common and most reliable probability selection is the simple random sample. Here, elements of the population are selected which all have the same probability of being selected for the sample. This can be done with the help of a random generator.

Stratified sample

In a stratified random sampling, random samples are drawn from different strata (e.g. income classes low, medium, high). Sampling can be proportional (the size of the shift sample is proportional to the size of the shift) or disproportionately.

Stratified samples are used when there is a high degree of heterogeneity with regard to a certain characteristic in the population. A prerequisite for stratified samples is that there is prior knowledge of the distribution of characteristics and of the subjects belonging to a specific stratum.

Lump sample

In multi-stage cluster sampling, the random selection extends over several stages, with whole units (clusters, e.g. school classes) of the population being selected in a first stage. In a second stage, elements of the respective units are then selected (students of the selected school class).

Conscious choice

Odds selection

The quota selection (quota sample), which is often used in market and opinion research, intends to define a sample that corresponds to the distribution of characteristics of the population. The quota selection is carried out according to certain rules, with the quotas representing a certain distribution of characteristics (for example 54% men, 46% women), which must be achieved through the sample selection. The respondent adheres to an exact quota, but chooses the respondents at his own discretion.

The choice of odds can lead to a number of distortions. An important one is the over-presence of certain, often easily accessible groups of people such as housewives.

In market and opinion research, the quota method is also used in combination with random selection (random quota). For example, municipalities are selected at random, and test persons within the municipality are then selected according to quota selection.

Snowball technique

In the snowball sampling technique, a small group of feature carriers is selected in a first step. In a second step, the contacts of those already questioned are then used to select further samples.

Arbitrary selection

Arbitrary sample

In the case of convenience sampling, carriers of features are included in the sample in an uncontrolled manner. They are taken into account because they are readily available or because they volunteer. Random samples are often found in psychological or medical tests or experiments. They can be used to test related hypotheses, but not to draw conclusions about the population.

Sample size

The definition of the size of the sample, given in absolute size and denoted by "n", depends on various determinants such as the available cost and time budget or the required precision.

Basically, the larger the size of the sample, the more precise the results of the sample will be. This raises the question of which sampling error should be tolerated. The smaller this tolerance, the larger the size of the sample must be defined.

It is advisable to consider the expected response rate at the beginning of the investigation. If, for example, results of around 450 people are to be available on the occasion of a quantitative survey and an expected response rate of 30% is expected, a sample size of around 765 people should be planned.

Ultimately, the question arises whether the selected sample and its size can solve the research problem to be worked on and thus meet the required representativeness for the research topic.

In addition, a selected sample size must be justified. It is important here that a representativeness is not postulated that does not exist.

Learning center Evidence of achievement Resources