4 SRM (Sampling) Flashcards
➢Sampling design process
1.Defining the target population
2.Determining the sampling frame
3.Selecting the sampling technique
➢ Esp. Non/probablity
4.Determining the sample size
5.Detailed instructions on sample selection procedures
6.Selecting actual elements
7.Carrying out the fieldwork
SAMPLE vs CENSUS?
❖Census – studying the whole population.
➢ If the population is relatively small (~200 or less) (usual in
organizational research, e.g. party members, MPs, etc.).
➢ If the variance is not distributed normally or is very large.
➢ No scarcity of budget and time.
❖Sample – subgroup of the elements of the
population whose statistics will be used to
estimate the parameters of the population.
❖When to use sampling:
➢If the population is large (usual in societal issues).
➢Budget and time constraints.
➢Destruction or contamination of the sample
(research requires consumption of the item under
investigation; need of secrecy).
➢Need for in-depth study of individual cases.
1st step of SAMPLING DESIGN PROCESS
- Define the target population precisely in terms of:
❖Elements
➢The object from which/ about which the information is wanted [compare to unit of
analysis] – usually respondents, in macro-level research – countries.
❖Sampling units [if applicable]
➢The basic unit containing the elements available for sampling.
➢E.g. if the desired respondents are heads of households, the sampling unit is
households.
➢If you have access to the target population directly, the sampling unit is the same as
the element.
❖Extent – geographical boundaries, locations.
❖Time – the time period under consideration.
2nd step of SAMPLING DESIGN PROCESS
- Determine the sampling frame
❖Representation of the elements of target population from which you’ll choose respondents.
❖An actual list or set of directions for identifying the target
population (e.g. the telephone book for telephone surveys; an
organization’s mailing list).
➢ If you cannot compile the full list, you need to specify directions
for choosing respondents.
➢ For location surveys (e.g. voters at the polling station), the
sampling frame is everyone who shows up there during a specified
period of time – so, essentially a census within a narrow frame.
❖If your list does not completely coincide with the population, you will
have a sampling frame error and your inferences will be inappropriate
(there will be a difference between the parameters of the population
and the statistics of the sample).
❖Possible ways to treat a sampling frame error:
➢Redefine the population in terms of the sampling frame.
➢Screen the respondents in the data collection stage to ensure they meet the requirements for the target population – this will eliminate inappropriate elements but will not account for omitted ones.
➢Adjust collected data by weighting.
Sampling and nonsampling errors
❖Sampling errors (difference between parameters
of population and statistics of sample).
vs.
❖Nonsampling errors (problems in data
measurement, collection process and analysis –
esp. dangerous with large samples).