Quartz 4

survey-nonresponse

3 min read

Type: course

Up: index

Course Description

Presentation

presentation-increasing-response-rates-incentives

Syllabus

  • Introduction, total survey error framework, developments of non-response rates
  • Non-response components, non-response in various survey modes
  • Response rates and non-response bias
  • Increasing response rates: advance letters and contact attempts
  • Increasing response rates: reminders
  • Increasing response rates: Incentives
  • Assessing non-response bias: follow-up studies
  • Assessing non-response bias: basic question approach
  • Preventing non-response bias: adaptive and responsive design
  • Compensating for non-response bias: weighting using socio-demographic
  • Compensating for non-response bias: weighting and paradata

0: Introduction

What is a survey?

2 important components.

Questionnaire

  • A set of questions
  • Standardized
  • Should be in fixed order because order of questions can effect

Sample

  • A sample should be random
  • We need a specific mechanism to select our respondents which is random

The Aim of the Survey

  • Purpose of a survey

    • We ask individuals, however we strive for scholarly knowledge regarding groups or societies
    • We are not interested in individuals (a survey is not a test or an assessment)

  • We survey samples of individuals, however we aim to compute estimates

    • Estimates of parameters

  • We are interested in aggregated information (characteristics) from our sample (mean, median, mode)

  • We are estimating parameters, this is the aim of the survey, we try to estimate what would be the result if we had asked the questioner to the whole population.

  • A survey is not a test. Why?

    • Test has the right answers but surveys do not.
    • Test can be used to evaluate the individuals, it may have consequences for the individuals that are
    • In surveys we never use the data to describe the individuals, this is even more important when we think that people may fear if their answers may have consequences for them.

1: Total Survey Error

  • Part of the errors in our data during either during

    • Measurement process
      • When respondents answer our questions, there will be errors or estimates or false answers
      • The mode of the survey - is it going to be on the phone, face to face, online? this also introduces measurement errors
      • Validity
      • Measurement Error
      • Processing Error
    • Representation process
      • Coverage Error
      • Sampling Error
      • Nonresponse Error
      • Adjustment Error

  • These 7 error comes in two different types:

    • Variance
      • Random error
      • Precision of estimate (confidence interval)
      • Reliability
    • Bias
      • Systematic error
      • Accuracy of estimate (point of estimate)
      • Validity

Then what?

  • We try to correct our responses and sample with techniques (weighting)
    • Weighting should enhance the quality of estimates, for example: to equally include different population groups
  • Sometimes erros can add up to each other, sometimes they can even or compansate for each other

Graph View

Backlinks