Thursday, February 10, 2011

Chapter 1: An Introduction to Survey Errors


Diverse Perspectives on Survey Research

-          Survey vs poll
-          Describers and modelers use survey data differently
o   Describers: discover some fixed property of a set of people (ie unemployment rate)
§  Does sample reflect population
o   Modellers – motivated by hypotheses about cause of social phenomena
§  Does survey measures fully capture concepts involved in their theory
-          Two types of error
o   One
§  Sampling error
·         Error that arises because only subset of population was measured.
·         Increase sample size
·         Assure representation of different groups in sample through stratification
·         Adjust probabilities of selection for different sub groups
·         Use clustered selections to reduce cost.
o   Two
§  Non sampling error
·         Reduce non response and measurement errors
·         Better worded questions
·         Interviewer training
o   Interviewer effect
·         Reduce / measure this error

The themes of integration: errors and costs

o   Costs of survey activities act as limiting influence on efforts to reduce survey error

The languages of error

o   Statistics
o   Psychology
o   Economics

Classification of error within survey statistics

-          Bias
o   Constant error
-          Variable errors
o   Values differ over units (ie sampled persons, interviewers used)
-          Errors of non-observation
o   Coverage – eg, telephone surveys do not include people without a phone
o   Non response – respondents in frame cannot be located / refuse
o   Sampling
-          Errors of observation
o   Interviewer
o   Respondent
§  Ie, not able to correctly remember
o   Questionnaire / instrument error
o   Mode of data collection
-          A more restrictive view, sampling statistics
-          Other error terms in survey statistics
o   Accuracy
o   Precision – converse of variance – stable over replications
o   Reliability – not same meaning when used in psychometrics – here means same as precision

Terminology of errors in psychological measurement

-          Notion of unobservable characteristic that is measured with survey indicator
-          Problem is is not impossibility of measuring the characteristic, but weakness of measure
-          Classical score theory
-          Validity and reliability
-          Estimating validity
-          Other notions of validity

Language of Errors in Econometrics

-          Language of estimation of GLM
-          Selection bias

Debates About Inferential Errors

-          How does researcher conceptualise impact of errors of non-observation
o   Modeler: concentrate on correct specification of form of model and less so on errors of non-observation.

Important Features of Language Differences

-          Sampler is committed to randomization process that on average produces samples with desirable properties / analyst is more concerned with ability of this single sample to describe population


Summary: The Tyranny of the Measurable



No comments: