Sampling Sampling •Objective •Size does matter •Sufficient for purpose •To produce as accurately as possible a population on a smaller [useable/fit for purpose] scale • Targeting participants • • Whatever method of data gathering is used a systematic consideration has to be made about who to interview, or send questionnaires to, what to observe, which documents to analyse, etc. • • • Targeting – 3 Ws –When data is collected – before, during, immediately after, sometime after or combination of these? – –Where data collection is done – what is best for participants to limit social desirability effects yet is practicable? – –What should information be gathered about – events, processes, self-reported behaviour / attitudes / beliefs, etc. • • Targeting •This means thinking about from whom/what information is gathered: •Are your participants’ views typical of a group? •Do they have specific kinds of information or perspectives which it are important to gather e.g. because they were happy or dissatisfied with a project? • SAMPLING •Of people •Of events •Of institutions •Of extant data •Of time •Of artefacts Population [or sample frame] •Used to refer to any social group or complete data set • e.g. from a whole society to employees of an organisation or the 2nd year students in ESF. • • Sample Size •Main parameters are: • •complexity of population • •research methods used • •time and cost factors • Sample Size • Attention has to be paid to resultant cell size (i.e. the subgroup of the population you end up with once you have applied sampling technique) for reality and statistical purposes. • • For example, in a typical education situation 100 pupils divided by gender and, for e.g. two levels of attainment gives cell size 25, any further variables reduces cell size by 50% etc. • • Reality: is sample big enough so that findings are generalisable in common sense terms – 80% = a majority but may only mean 4 people is the sample size is only 5! 4 people’s views do not make a sound basis for generalising to the parent population. • • Statistically: sample has to be big enough to make statistical testing a reasonable thing to do. • Non-respondents •Can be catered for by: • •Over-sizing the initial sample • •by additions or replacements Systematic Random Sampling •Time saving •take every nth number starting at random number, e.g. for 10 per cent sample 5/12/25 etc. •Must make sure there are is no pre-existing structure to the list that biases the sample…….for hypothetical example, if a list of married heterosexual couples went husband-wife etc and you chose every 10th person, every person in the sample would be female! Random Sampling •It must be established that the population/sample frame has no order/arrangement that will bias the research • •the frame may have an order that predisposes the sampling of a particular type. The phone book is in alphabetical order – but only lists those that have a phone and are not ex-directory etc. Would this bias the research? Stratified Sampling •Population/sample frame is first divided into strata from which a sample is drawn •Stratification is the process of grouping members of the population into relatively homogeneous subgroups before sampling. The strata should be mutually exclusive : every element in the population must be assigned to only one stratum. The strata should also be collectively exhaustive : no population element can be excluded. Then random or systematic sampling is applied within each stratum. This often improves the representativeness . Multi-Stage Sampling •Samples within samples or by stage, e.g. Schools in Britain, first sample LEAs, next schools, next classes, next pupils in classes. • •High level of sophistication can be achieved, e.g. in above first stage could include size/rural, urban location/per pupil expenditure, etc. Data can vary by stage Quota sampling •Non-random stratified sampling that is common in market research. A researcher collects data from a quota of individuals defined by gender/ethnicity/employment status etc. • •Dependent on accuracy of design and execution. Opportunity sampling •‘Captives’, like a class of students, club members, patients in a ward/clinic, customers in a shop etc. Volunteer sampling •By invitation, e.g. readers of a publication, victims of bullying/crime, users of a given commodity e.g. doctor’s surgery etc. • •Useful where population is disperse. • •Limitation is that the difference between volunteers and non volunteers might be crucial to the research. Response rates •Soundness of evaluation/research depends on response rate. • •Refusals and incomplete responses. • •Response rate can depend on interest of topic/instrument design/time to complete. Best and worst results •Best is usually face to face, on the spot • [90+% response] • •Worst is usually postal / email / web questionnaires • [40% considered very good] •