I would like to apply Confirmatory Factor Analysis (CFA) to a Likert-type questionnaire data. It is supposed that this data is affected by response bias: some patients either overestimated or underestimated their symptoms when filling up the questionnaire (which might be somewhat similar to acquiescence bias and its opposite). So, if an answer to each questionnaire item ranges from 1 to 5 (where 1 = “strongly disagree”, and 5 - “strongly agree”), some patients tend to give higher scores and some tend to give lower scores regardless or their actual symptoms. The question is how to correct for such bias and keep data meaningful for Factor Analysis? And in particular: What are possible approaches? How to choose among them? Is there a common approach?

  • 1
    $\begingroup$ How do you know there is such bias? $\endgroup$ – StatsStudent Jan 10 '19 at 15:45
  • $\begingroup$ @StatsStudent, our clinical experts strongly suggested to take it into consideration as such bias is common with these subjects. I am afraid I can't clarify it further. $\endgroup$ – Ilia Pershin Jan 12 '19 at 16:29
  • $\begingroup$ I'm afraid if you don't know the direction or degree of the bias and have not conducted any validation studies, it will be impossible to correct for such bias without conducting some sort of validation study to determine this. The good news is, in this case, it sound like you have a suspicion that the bias might be net zero in aggregate large samples since some overestimate while others underestimate. $\endgroup$ – StatsStudent Jan 12 '19 at 16:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.