I'm doing a correspondence analysis as part of a study. Unfortunately, one of the commands I use with R doesn't work because I have supplementary categorical variables that the package doesn't recognize/process. I have read Greenacre and his example on the issue here :
"The example of column Y and the vertex points in Exhibits 12.2 and 12.4 should not be confused with what is called \dummy variable" coding, a subject which we shall treat in detail when we come to multiple correspondence anal- ysis in later chapters. For example, suppose that we had a classication of the scientific areas into \Natural Sciences" (NS) and \Biological Sciences" (BS), the latter group including Biochemistry, Zoology, Microbiology and Botany and the former group containing the rest. A standard way of coding this in CA is as a pair of dummy variables, NS and BS say, zero-one variables with the values NS = 1 and BS = 0 for Geology (a natural science), for example, and NS = 0 and BS = 1 for Biochemistry (a biological science), and so on. One might be tempted to add these dummy variables as columns of the table and display them as supplementary points, but this would not be correct. This is not a count variable like the Y variable, which happened to have had 0's and 1's as well; in that case the data were real counts and could have been other integer values. The correct way to display this NS/BS information is as a pair of rows, similar to the way we displayed Math Sciences above. That is, sum up the frequencies for the NS rows and add an extra row called NS to the table, and do the same for the BS rows. In this way the NS and BS points will be weighted averages of the points representing the two sets of scientific areas (we shall return to this subject in Chapter 18).". (Greenacre, 2017, p.95)
I understand that in this case for example, the variable N.S could be coded 1 for the presence of "Biochemistry, Zoology, Microbiology and Botany" and 0 if one of the 4 is absent and that is what makes the dummy coding incorrect. On the other hand, do you object to dummy-coding each modality of this variable? Example: Biochemistry 1: present; 0: absent; Zoology 1: present; 0: absent etc ... Or does the problem still remain?
Ref : Greenacre, M. (2017). Correspondence analysis in practice. CRC press.