For example, respondent A may minimize things and so, his “agree” is really “strongly agree” where as respondent D tends to over-exaggerate matters and so, her “strongly disagree” is really “disagree.”
Assuming these differences cancel each other does not seem prudent, or I am mistaken?
Wouldn’t it be necessary to first establish each respondent’s baseline before asking opinion / subjective questions?
Note. my question is not about (re)coding i.e. Strongly agree => 1, …
Thanks in advance for any clarification or suggestions.
Yes, respondents’ scoring could be similar, but it would be dangerous to assume this is / will be the case. (Pls, refer to latest version of Practical Research by Ormrod & Leedy, page 144.) Unfortunately, they do not offer a way to account for or minimize this idiosyncratic variability between respondents.
In doing my literature review, I looked at several research questionnaires. All but one survey used 3 to 4 items to measure the construct. However, most of the surveys examined (not specifically search for) used a series of Likert-scale questions to measure a construct.
To me, this results in measuring a construct on multiple subjective dimensions.