5
$\begingroup$

I want to assess item-total correlations on a 19-item questionnaire (some of the questions are meant to be reverse-scored). My question is:

  • Do I reverse score the items PRIOR to calculating the item-total correlations (in order to eliminate any variables that do not correlate with the total at >.40)?
  • Additionally, should the items be reverse-scored prior to running a factor analysis?
$\endgroup$
4
$\begingroup$

Yes, you should reverse score all items as needed to ensure that a particular score means the same thing on all items. You should do this for all types of analysis.

For example, you have 'propensity to shoplift' measured via 3 items on a scale of 1 to 5 (where 1 is low propensity to shoplift and 5 is high). Suppose that you reversed item 1 on the on the survey so that 1 is high and 5 is low. Then you should reverse the score for item one so that 5 means the same thing across all three items (i.e., 5 is high propensity to shoplift).

$\endgroup$
7
$\begingroup$

Reliability Analysis: Yes, you should reverse score the reversed items.

Factor Analysis: It does not matter so much. Eigenvalues and associated indices (e.g., variance explained by factors, rules of thumb regarding number of factors to extract, etc.) should be the same. The sign of factor loadings will flip based on whether you reverse reversed items.

$\endgroup$

Your Answer

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