I have a number of group variables that I capture on MCA plots. I also obviously have the coordinates for these plots in the X, Y and Z dimensions. i.e, most people plot the maps as 2 dimensional and then descriptively outline what is going on with the analyses.
You can also do MCA plots in a 3rd dimension (Z dimensions). This would be a 3D plot
Obviously MCA is a visual representation. I was wondering if there are any recognised tests for comparing group differences (e.g, male and female, or 3 way group splits) in the 2D and 3D cases.
I asked Greenacre and received no reply as yet. I cannot find literature here answering this condition.
Brad
Hey Bradley,
I guess the easiest was to compare groups is a t-test. An ANOVA or multiple regression can also be used to compare groups. I hope this helps.
Best regards
Dear Paul et al,
Thank you for your considered reply. Yes the data set is from one sample of which Male and Females responded to the same image/attribute dimension question. The study is actually around brand personality question items.
You are right in saying that the combinations of response. could be male/ HI or LO income, Female/HI or LO income.
I can see what you are saying regarding nonlinear analyses and plots.
However, my question pertained to Multiple Correspondence Analysis..... and lets say you had the two datapoints calculated and plotted for the MCA plot. Is there any way of calculating that the two plotted points in for the two group are different?
I am not sure whether I misinterpreted your reply BUT it did not appear to be in a MCA framework.
best regards
Brad
Dear Paul,
I conversed with a co-author and expert in MCA. He wrote a number of then DOS programs in the 1980s that are still used by Market researchers today. I think aspects of his response mirror your response BUT he talks about things less spatially.
I will post this reply in the next day or two
regards
Brad
Dear All,
I said I would pick up this theme again. A co-author/mentor states that, "But you cannot test coordinate point positions in isolation (referring to an MCA plot/points). And we have to go back to the numbers - there are no statistical tests involving geometry (although apparently the great Fisher used visual spatial patterns to think through his stats ideas!) . When something is mapped in MCA the whole thing is connected.
If it is only two points one still has to look at the whole. One would be looking at a contingency table - but a multidimensional one. There is a whole way of testing these based on logit models if I recall rightly. Horrible stuff and one of the theory problems was around sparse matrices. This was because once you got up to 4 and 5 dimensions the greater the cells and many become pretty empty."
I think this is similar to the statements from Paul in nonlinear indexes.
I am not sure I have my head around the issue completely.
I look forward to the replies.
Brad Wilson
Ok. let me read a second time...
Would you agree with the contingency table approach that was talked about?
brad
I will get through the readings by end of next week. I thought this question may have unearthed a few more spatial mapper types!
I might take you up on the gratis copy of your software. should I send you personal message
regards,
brad
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Computer Graphics and Animation