Sampling

Sampling

1. Sample or census

   • A census contains complete enumeration of the entire population
     • Unrealistics if the population is large
     • Can increase non-sampling errors
   • A sample is a sub group of the population selected
     • Budget and time limits are constraints for using sampling
     • Good for focusing on an individual case

2. Sampling design process

   • 1) Define population
   • 2) Determine sampling frame
   • 3) Select sampling technique
   • 4) Determine the sample size
   • 5) Execute the sampling process

3. Two sampling procedures

   • Non-Probability
     • Personal judgement
     • Researcher can decide what is in sample
     • Do not allow objective evaluation of precision of sample results
     • Commonly used in non- probability sampling methods
   • Probability
     • Selected by chance
     • Fixed probailistic chance of being selected
     • Requires precise definition of target population

4. Types of NON probability sampling

   • Convenience
     • Right place at right time
     • Cheapest and least time consuming of all techniques
     • However they do not represent any of the definable population
     • Not recommended for descriptive or casual research
     • Examples = students, church groups, street interviews
   • Judgemental
     • Form of convenience sampling
     • Judgement over expertise
     • Inexpensive, convenient and quick
     • Does not allow generalisation
     • Subjective and the value is dependant on researcher's judgement
     • Examples = Product testing with people who are fussy
   • Quota
     • Two stage restricted judgemental sampling
     • First stage = developing control characteristics
     • Second stage = Sample elements are selected based on convenience or judgement
     • Freedom to select elements providing they fit the controlled characteristics
     • A good quota must ; specify characteristics and know the distribution of the characteristics
   • Snowball
     • Initially a small group is selected sometimes on a random basis
     • Respondents are then asked to identify others with similar interests etc.
     • Referrals from referrals = Waves
     • Probability can be used to initally select respondents, non probability is the final outcome
     • Main aim is to estimate characteristics that are rare

5. Types of PRObability

   • Simple random
     • Equal probability of selection
     • Each element must be enumerated
     • Random numbers are used to select respondents
     • Easily understood and can be projected to a whole population
     • However they can be difficult to construct as the geographic spread may be non representative
   • Systematic
     • Selecting random starting point
     • Population size (N)/ sample size (n) = i
     • Researcher assumes there is order
     • Less costly
     • Can be used without knowledge of the elements
   • Stratified
     • Two stage process
     • Population is split into sub groups or strata
     • Elements are selected based upon probability rather than convenience or judgement
     • Criteria for selection = homogeneity, hetrogeneity, relatedness and cost
     • Samples can be proportionate or disproportionte (cluster)
   • Cluster
     • Two- step process
     • Divided into exclusive sub pops or clusters
     • Elements are then randomly drawn from clusters
     • May be proportionate or disproportionate

6. Adjusting for nonresponse

   • Subsampling of nonrespondents
     • Contacts a sub sample of non respondents
     • Results in high response rate
     • Values obtained can then be projected to all non respondents
   • Replacement
     • Replaced with non respondents from other similar surveys
     • Can offer incentives
     • Important that the two surveys are similar
   • Substitution
     • Subgroups used to identify substitiues that are similar to particular non- respondents
   • Subjective estimate
     • Evaluate the likely effects of non-response based upon experience
     • Information can provide a basis for evaluating the effects of non-response
   • Trend analysis
     • Trend between early and late respondents
     • Can estimate where respondents stand on the character of interest
     • Projected technique
   • Weighting
     • Attempts to account for non-response by assigning differential weights to data
     • Can destroy self- weighting nature of sampling design
     • Can introduce complications