survey notes
week 1; errors — inference 1
survey
survey research is about producing statistics and drawing inferences
about a target population
advantages disadvantages
relatively low costs generally limited to scaled or check-list questions
relatively fast delivery no control over response rate
rapid data processing provides information more than understanding
can research large populations increasing public resistance
various modes: phone, mail, internet, face-to-face difficult to explore issues in depth
two inferences
we should be able to measure a certain attitude if we use proper questions
if we pull a proper sample we should assume the sample is representative to the population
survey notes 1
, errors
common sources of error in survey:
poorly worded questions
deviations from the script by the interviewer, mostly with face-to-face or phone
interviews
misunderstanding on the part of the interviewee
memory problems on the part of the interviewee
the way the information is recorded by the interviewer
the way the information is processed, either when answers are coded or when data are
entered into the computer
two types of error
random error: affects findings in unpredictable ways
systematic error: affects findings in predictable ways
error in sample: sample differs from error in question: answers don’t reflect variable
population
1. random = question invalidity
1. random = sampling error
differs only by chance = random
difference only by chance = variability / honest mistake in questions
random variability
2. bias (systematic) = question bias
2. biased (systematic) = sampling
social desirability bias
bias
can lead to overestimation of
the sample
only certain groups respond
question bias
survey notes 2
, difference in what is asked
objective facts
can be biased in one direction
error/validity can be measured bias
subjective states
can not be biased
validity can not be measured
—> validity = difference between answer and true score
level of bias:
bias at respondent level —> only for one
bias at aggregate level —> for everyone
designing questions
general rules
remember your research question
decide exactly what you want to find out
consider your audience
imagine yourself as a respondent ( would you answer the question?)
common mistakes
ambiguous terms ( often, frequently)
long questions
double barrelled questions
survey notes 3
week 1; errors — inference 1
survey
survey research is about producing statistics and drawing inferences
about a target population
advantages disadvantages
relatively low costs generally limited to scaled or check-list questions
relatively fast delivery no control over response rate
rapid data processing provides information more than understanding
can research large populations increasing public resistance
various modes: phone, mail, internet, face-to-face difficult to explore issues in depth
two inferences
we should be able to measure a certain attitude if we use proper questions
if we pull a proper sample we should assume the sample is representative to the population
survey notes 1
, errors
common sources of error in survey:
poorly worded questions
deviations from the script by the interviewer, mostly with face-to-face or phone
interviews
misunderstanding on the part of the interviewee
memory problems on the part of the interviewee
the way the information is recorded by the interviewer
the way the information is processed, either when answers are coded or when data are
entered into the computer
two types of error
random error: affects findings in unpredictable ways
systematic error: affects findings in predictable ways
error in sample: sample differs from error in question: answers don’t reflect variable
population
1. random = question invalidity
1. random = sampling error
differs only by chance = random
difference only by chance = variability / honest mistake in questions
random variability
2. bias (systematic) = question bias
2. biased (systematic) = sampling
social desirability bias
bias
can lead to overestimation of
the sample
only certain groups respond
question bias
survey notes 2
, difference in what is asked
objective facts
can be biased in one direction
error/validity can be measured bias
subjective states
can not be biased
validity can not be measured
—> validity = difference between answer and true score
level of bias:
bias at respondent level —> only for one
bias at aggregate level —> for everyone
designing questions
general rules
remember your research question
decide exactly what you want to find out
consider your audience
imagine yourself as a respondent ( would you answer the question?)
common mistakes
ambiguous terms ( often, frequently)
long questions
double barrelled questions
survey notes 3
