class 1-TSE overview Flashcards
Demming (1994)
first listing of sources of problems-beyond sampling issues
classification of factors affecting the usefulness of a survey
-bias of the auspices
- bias from late reports
- bias from unrepresentative selection of the date for the survey
Total survey error
an indicator of data quality measured by the accuracy of the mean square error of the estimate
Demming (1994)-errors
nonresponse sampling interviewer effects mode estimation step errors (wrong weighting) NO COVERAGE ERRORS- drawing a sample in a population
Kish (1965)
-focus on biases
separates errors of observation and non observation
-does not note that sampling errors are errors of non observation
analytic errors
emerging research
estimation step errors
Types of non sampling biases
non observation
observation
types of Sampling biases
frame biases
consistent sampling bias
constant statistical bias
Mode error bias
differing responses based on mode of administration
e.g., variance- answer quickly in internet mode
Four components of error
frame errors-missing elements, improper use of clustered frames
sampling errors
non response errors- item non response, unit non response
measurement errors
sources of observation errors
field-data collection office-processing interviewer respondent instrument mde
sources of non observation errors
non coverage
non response
sampling
Bias
systematic differences between expected value of survey estimate and population value
non observation errors
departure of survey estimate from pop value because of failures to measure some pop measure
Specification error
“construct validity”
Do we measure what we planned to measure
observation errors
departure of survey estimate from pop because of deficiencies in the measurement process
Total Survey Design- Dalenius (1974)
Requirements-everyone needs to respond
Specifications-contact households in person
Operations- fieldwork stops after 5 attempts
Groves (1989)
linkage between TSE and psychometric true score theories and measurement error and selection bias
Intro to Survey Quality (2003) Biemer and Lyberg
division of sampling and non sampling error
adds specification error
fitness of use- usability of surveys
variance
dispersion (over replications) of departures between survey estimate and expected value of survey estimate
total survey error evolution
began as a way to caution against focusing on sampling error only
contains statistical and non stat notions
1970s- attention to total survey design
1980-1990s- include psychometric
understudied in TSE
Specification error processing error bias linkage errors estimation errors
weakness of TSE
Exclusion of key quality concepts -a user perspective lack of routine measurement ineffective on proff standards burden on design of some estimators
Indicator quality
credibility
relevance
estimator quality
data quality
credibility
fairness
independence
transparency
integrity
relevance
promptness
completeness
effectiveness
estimator quality
accuracy
revisability
data quality
reliability
validity
confidentiality
response burden
Strengths of TSE
Separation of whats affecting stats differently
conceptual foundation of survey meth
able to identify problems in research
TSE and costs
efforts to reduce errors increase costs
