I am running an EFA on an 8-item questionnaire. Each item can be rated between 0 and 10, with higher scores = more negative/worse relative to lower scores. Sample size is small at n=81. I am treating the data as ordinal, so am looking at polychoric correlations. I have done some preliminary analyses but have struck problems as item 7 on the scale has missing response categories (no participants chose 6, 7, 9 or 10) and so polychoric correlations could not be calculated. Must I remove this item for the analysis to work or are there other ways to manage this?
Employ listwise computation or pairwise deletion by cleaning and removing the offending items, this could be tedious in large item sets.
Prudence Butler Which software are you using? As far as I know, Mplus can handle ordinal variables with different numbers of categories in the same analysis using polychoric correlations. You might want to check this out.
If your scale ranges from 0-10, you are probably fine to run Pearson's correlation.
Thanks all for your suggestions. Christian Geiser I used an online program called FACTOR (see https://psico.fcep.urv.cat/utilitats/factor/MainMenu.html). I'll see if I can get access to MPlus.
D. Eastern Kang Sim could you point me in the direction of any references that indicate that it's ok to treat 10 point Likert scales as continuous? I found this paper (Article Can Likert Scales be Treated as Interval Scales?—A Simulation Study
) but perhaps there's better ones. I can see that choosing whether to treat Likert scale responses as ordinal Vs continuous is an ongoing debate!- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size