5
$\begingroup$

I have a set of 20 participants for a study who were given a 7-point Likert scale that was made incorrectly by a graduate student. The student accidentally did not add the 6th answer level. The scale for these 20 participants arguably has no validity now. However, I am hoping there is a way to use it for a pilot or preliminary poster. The study will continue for a couple years and is proposed to have 300 participants. But I would like to use these 20 participants for a small pilot or preliminary poster. Is there a way for me to still use that scale (missing the 6th answer option) and the 20 participants?

I don't know if I can impute it because the data is not MCAR, the answer level was not added to the scale as it should have been. But can I weight the low and high-ends of it or something still? I would like to see if the measured construct from that scale relates to either another variable (such as Engagement) or the outcome variable. It is a longitudinal study, they come in once a week for three weeks. Each time they come in to the lab and use the online intervention they take the scale three times (i.e., Baseline, after first half of website, and then after second half of website), which means we have each participants using the scale 9 times.

Improve this question
$\endgroup$
4
  • 1
    $\begingroup$ Please confirm that with "the 6th item" you do not mean the 6th question of a multi-question test (as MariaAnt suggests in her answer below) but rather the 6th category label on the Likert scale. Example: Likert scale offers 7 choices: "never, rarely, sometimes, frequently,often,mostly,always" and "mostly" was omitted. $\endgroup$
    – matus
    Commented Sep 13, 2015 at 10:07
  • $\begingroup$ Correct, by item I do NOT mean the 6th question. All of the questions are still on the scale, it is missing the 6th item. $\endgroup$
    – Austin
    Commented Sep 13, 2015 at 14:40
  • 1
    $\begingroup$ I didn't get that. The term "item" is usually used to refer to a question or statement within a scale. You are referring to the number of "options", or "levels" of the scale en.wikipedia.org/wiki/Likert_scale $\endgroup$
    – user7759
    Commented Sep 15, 2015 at 9:56
  • $\begingroup$ I found AM's question interesting and investigated it in more depth. My investigation is not suitable for the format of stackexchange but may be nevertheless of interest, so I just link to a post on my blog. I hope it's ok if I leave it here in the comments. $\endgroup$
    – matus
    Commented Oct 30, 2015 at 14:51

1 Answer 1

Reset to default
3
$\begingroup$

I think you'll come unstuck if you don't have a balanced scale.

If you consider a typical five-point scale:

  • Strongly disagree
  • Disagree
  • Neither agree nor disagree
  • Agree
  • Strongly agree

If we forget to put Strongly disagree on the scale, it is positively biased, because it is heavier on the Agree side of the scale. If we forget to add Strongly agree we have a negatively biased scale.

If you intend the questionnaire to be used for research purposes, you'll need to ensure the scale is correct - otherwise it may affect your findings, and it will definitely cause people to question whether the outcome would have been the same as if the correct scale was used.

Improve this answer
$\endgroup$

Not the answer you're looking for? Browse other questions tagged or ask your own question.