I have done my survey, and now I need to do the data analysis. However, I am not sure whether or not I need to do the reverse code at the first step before I can do the remaining analysis.

There are 5 constructs in my questionnaires with 5-point scale (i.e. strongly disagree, disagree...) and now my question is all the questions in each construct don't have a negative sentence, I mean, for example: I like playing football in the rain (strongly disagree, disagree, agree, neutral, agree, strongly agree) So, in this case, should I reverse code before running the analysis?

  • $\begingroup$ if by 'reverse code' you mean 'rescaled' so that a single direction of response (e.g. 'I agree that I like' and 'I disagree that I dislike') can be compared across questions, then probably yes. $\endgroup$
    – katya
    Mar 2, 2016 at 4:04
  • $\begingroup$ reverse score for Cronbach Alpha for positive questions $\endgroup$
    – thavi
    Mar 4, 2016 at 5:35

1 Answer 1


You can do the factor analysis with items that are scored in different directions - some loadings will just be negative and some positive. The signs of the factor loadings are arbitrary. More of a latent variable is less of the opposite latent variable - more intelligent = less stupid. More happy = less sad etc.

However, it is common to do Cronbach's alpha on factors and that will be messed up by reverse coded items. It may also be easier to interpret the factor analysis results on items that are all in the same direction.

  • $\begingroup$ thank you. Should i do the reverse coding before performing the Exploratory Factor Analysis even though my questionnaires are in positive sentences? $\endgroup$
    – thavi
    Mar 3, 2016 at 11:35
  • $\begingroup$ It's really up to you, but I think it makes things eaiser. $\endgroup$
    – Peter Flom
    Mar 3, 2016 at 11:42
  • $\begingroup$ Should I reverse score before running the Cronbach Alpha when all my questionnaires are all phrased positively? $\endgroup$
    – thavi
    Mar 4, 2016 at 5:31
  • $\begingroup$ It doesn't matter how the questions are worded it matters how the questions relate to the latent variable. So, suppose you ask people to agree/disagree with the following two statements: 1. "I would never vote for a Democrat" and 2. I always vote Republican. These go in the same direction. $\endgroup$
    – Peter Flom
    Mar 4, 2016 at 19:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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