There are factor analytic methods (both EFA and CFA) that can handle a combination of binary (dichotomous) and ordinal (Likert) variables and that are implemented, for example, in the Mplus and lavaan programs for structural equation modeling. These methods use tetrachoric/polychoric correlations and weighted least squares estimators (referred to as WLSMV or DWLS).

Assuming that your multiple choice questions can be scored dichotomously (correct vs. incorrect), these could also be used.

Hello all,

IMHO, if I begin with defining/conceptualizing a latent variable as dimensional, I cannot really understand what the value is in presenting yes/no-indicators for the latent variable....

Best

Holger

hello,

Factor analysis can be applied on the all types of questions.The aim is to summarize and reduce responses.So it has no relation to the nature of the question.You can refer to chapter 14 of Thietard et al.for more in-depth information.

Hello Laila,

I agree with the second part of your answer. Regarding the first, may it be that you confuse that with PCA (principal component analysis)? These two are often used synonymously. Factor analysis specifically is a *model* that tries to identify underlying common causes of the observed variables--not to reduce the number of items. This is the goal of PCA (it applies no model but is a purely technical data reduction techique resulting in the lack of any surplus meaning of the components (not factors).

Best,

Holger

Apart from the answers received earlier, you may consider the following:

1. I do not see how your second option (Questions with multiple choice options "where one is correct) can be subjected to FA.

2. From the viewpoint of capturing more variation, in general, I'd prefer a Likert-type response.

Best,

Rajiv