1
$\begingroup$

I want to conduct an exploratory factor analysis on a small questionnaire that I have. The questionnaire consists of 20 items (N=100) that are scored on a 1-5 Likert scale (strongly agree - strongly disagree). I want to perform an EFA with polychoric correlations (principal axis factoring as the variables are non normal), parallel analysis to obtain the amount of factors, and oblique rotation (as the variables are almost always correlated).

However, besides the 1-5 Likert answering options there was also an option "(0) does not apply". I have read numerous books/articles on the subject, but can't seem to find an answer on what I should do with the 0's? Should I use listwise deletion (this will result in a smaller sample size) or should I set the 0 to missing? I want to use FACTOR for the analysis, they have an option for missing data using hot deck imputation. However, I don't know if changing the 0's to missing is ill-advised.

I hope someone has an answer for me. Anyway, thank you in advance for your valuable time.

Pcdl.

Fabrigar, L. R., & Wegener, D. T. (2011). Exploratory Factor Analysis (1 edition). Oxford ; New York: Oxford University Press.

Gaskin, C. J., & Happell, B. (2014). On exploratory factor analysis: A review of recent evidence, an assessment of current practice, and recommendations for future use. International Journal of Nursing Studies, 51(3), 511–521. https://doi.org/10.1016/j.ijnurstu.2013.10.005

Norris, M., & Lecavalier, L. (2010). Evaluating the Use of Exploratory Factor Analysis in Developmental Disability Psychological Research. Journal of Autism and Developmental Disorders, 40(1), 8–20. https://doi.org/10.1007/s10803-009-0816-2

$\endgroup$
3
$\begingroup$

If it makes sense for someone to have "does not apply" as a response but have a meaningful value for the factor, then maybe your factor isn't well-defined for all people. For example, if your proposed factor is "stress at home" and one of the questions asks "How much do your children cause you stress?" people who don't have children can only respond "does not apply", but if this question is relevant to the factor, then the factor is only well defined for people with children. The meaning of a factor comes from the items used to construct it; if an item does not apply to a group of units, nor does the factor. The factor must be redefined either to apply only to a subpopulation for whom all the items or relevant or to take its meaning only from the items that apply to all units in your population.

If you think 0s actually mean 1s (e.g., "How often do you smoke - 1 = rarely, 5 = often" and the participant doesn't smoke at all) because the response options were not specific enough at the low end, then recode them as 1s with good justification or add 0 as an additional response option in your analysis that takes on the value less than 1 (i.e., what the number 0 actually does).

If there is very good reason to have "does not apply" but the factor still applies to the individual and 0 doesn't mean less than 1, then you should treat it as missing and use the best practices for EFA with missing data. But to me, this points to a problem in your data collection and measurement than simply a quirk of your data that can be fixed by analysis.

$\endgroup$

Your Answer

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

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