Samples and Sample Size
For most experiments it's not possible to take measurements from every individual in a population either in a laboratory or in the field, as the full population is often too large. So a sub-set or sample is used to estimate the values that might have been obtained with a much bigger sample size.
A sample is made up of a series of sampling units which depend on the type of variable being measured; for example, a data value for a particular variable (eg. length) recorded for an individual sampling unit (eg. a plant) in a sample of n units (eg. n = 100) taken from the population under investigation (eg. plants in a field)
Sampling should allow sufficiently reliable information about the particular population under investigation.
The type of sampling or rather where the sampling occurs determines how much control the investigator has over the study. If the sampling is to take place in the laboratory then the investigator has much greater control in the design of the experiment. Design is critical in determining the sampling strategy and analysis of results.
When estimating population parameters from sample statistics, the sample size is important; larger sample sizes usually result in greater statistical reliability. However optimum sample size is a balance between statistical and practical considerations.
In deciding on appropriate sample sizes or checking whether available sample sizes are adequate, ethical considerations in experiments involving human and animal subjects should be observed.
Stratified sampling
This occurs when random sampling may still not be representative enough of the population and could miss out on certain areas. It is also useful in habitats that have clearly defined micro-habitats. Sampling occurs within sub divisions of the population or habitat
Example: in a population of organisms it may be necessary to sample within gender and age classes to get a more representative sample of the population.
Example: on a tidal shore stratified sampling may occur. Once the different zones from low tide to high tide are identified then sampling would take place randomly within the different zones present.