4.4.1 Sampling

• Sampling


  Examining a subset of the available data, known as the sample, instead of the entire data set
• Sampling


  • Used for a wide range of scientific, political and economic purposes for more than two centuries
  • Allows researchers to determine patterns and trends without examining the entire data set
• Sampling


  • Political opinion polls
  • Monitoring car journeys to study traffic congestion
• Representative sample


  A sample that accurately reflects the wider population in terms of relevant characteristics
• Creating a representative sample is incredibly hard, as the sample can contain hidden biases or omissions that will lead to mistaken conclusions
• Failure of opinion polls to predict 2015 UK general election


  • Polls systematically overrepresented Labour voters at the expense of Conservative voters
• Sampling has the drawback that both the sample and the data obtained from the sample are defined at the start
• Limitations of sampling in opinion polls


  • Landline-based polls exclude younger voters without landlines
  • Internet-only polls favour younger voters
• Pollsters cannot ask questions that are not in the poll, even if they realise a key question is missing
• Sampling in scientific investigation


  • Allows researchers to examine a tiny subset of the possible data due to time and expense constraints
• Genetic testing by 23andMe


  Samples a relatively small number of genes known to be associated with certain traits and conditions, to keep costs down
• Sampling a small number of genes


  Means diseases associated with genes outside the sample, or those whose genetic origin is uncertain, cannot be detected
• Analysing complete DNA sequences


  Avoids the problem of limited sampling, but is computationally intensive and more expensive