0

Is it possible to make Multiple Correspondence Analysis (MCA) with nominal data (such as country or gender) ? And more broadly, what are the assumptions of MCA?

For me, MCA is a type of factor analysis (FA) that allows to use nominal or ordinal data. But in this thread it is said that the first assumption of factor analysis is to have scale (interval or ratio) input variables. How MCA can be a type of factor analysis (that requires scale input variables) but also allow nominal or ordinal data ?

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
Siva Kg
  • 23
  • 3
  • 1
    Rather than saying "it's a type of factor analysis" it would be better to say something like "it is like factor analysis in some sense, but allowing for nominal/ordinal data". – Christian Hennig Jun 15 '21 at 13:50

1 Answers1

1

Yes, MCA is specifically useful for nominal and ordinal level data. According to this article, there are no underlying distributional assumptions.

mkt
  • 11,770
  • 9
  • 51
  • 125
T. C. Nobel
  • 111
  • 2