#AAPOR Presidential Address: The total error approach by Paul J. Lavrakas #MRX

AAPOR… Live blogging from the AAPOR beautiful Boston, any errors are my own…


  • “Applying a total error perspective for improving research quality in the social, behavioural, and marketing sciences”
  • Introduction; He grows orchids, collects movie dvds, and likes sci fi movies. [just as every president should!]  Core values – honesty, integrity, fair, dedicated to family profession and AAPOR
  • Much research is poorly conceptualized and interpreted; most studies could be improved with few if any cost implications
  • Total error approach can help do this, many are unfamiliar with this approach
  • [Yet another Bob Groves mention. this guy must be a hollywood celebrity :)]
  • “Survey costs and survey errors” 1989, Bob Groves
  • Total error – all problems in a study, all conclusions drawn that are wrong, anything that causes a study to be questionnable or limited in value
  • A way to plan the research, oversee data collection, interpret and disseminate findings, a rigorous process of self-evaluation and improvement
  • Bias – directional error, e.g, too high, too low
  • Variance – nondirectional error, imprecision, lowers confidence
  • Qual and quant researchers need to consider this
  • Representation side of research – have a target population, represent it through a sampling frame, leads to a designated sample the one you start with, end up with a final sample
  • Measurement – start with a construct, operationalize with measurement, gather responses, create a final dataset
  • Put representation and measurement together for final results and conclusions
  • Two top reasons for non-response error- we don’t ask them and they refuse
  • [long chat about different types of measurement error, read your intro to surveys text book]
  • Inferential error: when researchers use inappropriate or wrong analytic tools. Also drawing inferences that aren’t supported by the data
  • Consider terminology that applies to qual and quant – credibility, analyzability, transparency, usefulness
  • Credibility: scope – coverage error, sampling error, unit nonresponse error – external validity; specification error, instrument error, respondent error, interviewer error, construct validity
  • Analyzability: completeness and accuracy of analysis and interpretations, processing error, adjustment error, inferential error, peer debriefing, reflexive journals, triangulation, cause and effect reasoning
  • Usefulness: “do something” with the outcomes, support, refute, refine, generate hypotheses
  • [Reminds me of meta-analysis which only use published data which is generally only the interesting findings not the everyday more common boring findings]

One response

  1. “Yet another Bob Groves mention. this guy must be a hollywood celebrity” 🙂

    We definitely have our celebrities!

%d bloggers like this: