Multiple Factor Analysis (MFA) using FactoMineR

[@nicolaspinzon8181]

Simply great! thanks for taking the time to prepare these videos!

[@brijeshjani7997]

Thanks so much for the tutorial on MFA François!! I'm still trying to fully understand some basic concepts that go into setting up this MFA. For instance, what makes the smell, smellAf, tasting, and vis groups active and what makes the orig and pref groups supplementary. I was under the impression that active variables are those that a researcher actively controls, but aren't the smells/tastes/etc. all subjective data points? Any and all help is much appreciated :)!

[@janderz8]

Thanks for putting together these fantastic resources! Is it possible to get bootstrap reliability estimates for these sorts of models from your package?

[@HussonFrancois]

No this is not possible with the package to do bootstrap. But note that there is not an explicit model. It is rather a multivariate method to visualize data.

[@janderz8]

Thanks for the clarification!

[@galaxis38]

To confirm, the dataset has to be organized so that all the variables in a group are located next to each other in the data set, as in the picture on your slide 2, correct? The data set has to have it's columns organized by the groups?

[@HussonFrancois]

Yes you are right

[@galaxis38]

@@HussonFrancois Thanks for confirming!

[@cm8818]

Summary(res) is producing an error message. Why is R s frustrating?

[@galaxis38]

I keep receiving the following error: Error in MFA(dat, group = c(1, 10, 15, 3, 5, 9, 6), type = c(rep("n", : factors are not defined in the qualitative groups. My data set has 3 quantitative variables (currently class is integer) in one group from which I removed NAs by converting them to 0. I have 6 other groups which are qualitative categorical (one of these 6 is just a household id and is supplementary) and all are coded as 0 (no) or 1 (yes) (though sex is 0 female 1 male). So when I try to fix the error I have listed first, by turning the whole dataset into factors (except the 3 quantiative variables), this doesn't appear to fix the error. It gives me back this error again: Error in cov.wt(cbind.data.frame(res.ind.partiel[[group.actif[g]]]$coord.sup[1:nb.actif, : 'x' must contain finite values only. But there are no NAs or infinite values. Does MFA not like 0?

[@HussonFrancois]

You should consider the Id variable as the label of the individuas. When you read the data, use the argument row.names=1. Maybe it will fix the problem.

[@galaxis38]

Thanks for the suggestion! I was able to correct all errors and issues and get it running.

[@Rocadeluna]

Hi François! First of all, thanks! All the videos have been really helpfull. I wanted to ask for some help. I'm trying to perform an MFA but I keep getting this error: " Error in colSums(wt * y) : 'x' must be an array of at least two dimensions". I'd appreciate if you could give me some advice. I have looked for this error but the solutions haven't helped me. Thanks!

[@HussonFrancois]

Hi Paula, It is difficult to answer you but most of the time, this error appears when there is a problem in the dataset, such as some variables that are not read in a good way, or variables that are constant (the variance is equal to 0 and thus the variable cannot be standardized). You can send me your dataset (suppressing the names of the individuals and of the variables if you want) as well as your lines of code. FH