Saturday, February 19, 2011

Chapter 10: Measurement Errors Associated With the Questionnaire


 

-          Words in question

-          Structure of question

-          Order or context of the questions

 

-          Problem with survey measurement à respondent (object being measured) must itself be actively engaged in measurement process.

 

 

Properties in Words in Questions

 

-          No meaning can be given to a word, apart from cues from its context, by the respondent

-          A word can be taken to mean different things by same respondent.

-          Word can be taken to mean difference things by different respondents.

 

-          Example

 

o   Variability in meanings over different respondents who chose same answer

§  Pretty often

o   Context  of question may affect attributed meaning of response category

§  Meaning of word few

·         Reading a few poems before lunch

·         Writing a few poems before lunch

§  Meaning of word weekday

-          Rules

o   Respondents interpret interview rules to require them to answer each question, not merely to answer the questions they view as completely unambiguous

o   When faced with several possible meanings for a question, respondents will chose one for which an answer is more readily formed.

-          Respondents will answer survey questions despite little understanding of the question

-          Investigate meaning of different words in questions by

o   Split sample comparison

§  Do you think the US should forbid public speeches against democracy?

§  Do you thing the US should allow public speeches against democracy?

 

 

Properties of Question Structure

 

-          Length of questions

-          Open vs closed questions

-          Number and order of response categories in fixed alternative questions

o   Eg, number of scale points

o   Offering "about average" as an option

o   Alter 7 point fully labeled scale question into two questions

-          Explicit "Don't Know" response categories

-          Question structure in measurement of sensitive topics

o   Loading à behaviour is exhibited by many people

o   Randomized response technique

-          Question structure and recall of past events

o   Bounding interview

 

 

Properties of Question Order

 

-          Context effects on responses

-          Fatigue effect

-          Position of question in questionnaire

 

 

Conclusions about Measurement Error related to the Questionnaire

 

-          Stability of measurement is enhanced with many indicators of underlying concept.

 

 

Estimates of Measurement Error from Multiple Indicator Models

 

 

Cost Models Associated With Measurement Error through The Questionnaire

 

 

No comments: