MFA - Automatic description of axes - error: missing value where TRUE/FALSE needed

[Robrecht Bollen]

Dear FactoMineR community,

I am performing MFA with Factoshiny using sensory data from coffee evaluation. The data contains both cupping scores as well as sensory descriptive data. 

When I run the MFA in Factoshiny the graphics and analysis are executed, however the "Automatic description of axes" results in an error:

error: missing value where TRUE/FALSE needed

Is this missing value in relation with the dataset or a missing value in the analysis? How can I track back which values is missing?

Kind regards,

Robrecht Bollen

[Francois Husson]

Do you have the latest version of FactoMineR?
Or maybe you have a categorcial variable with as many categories as the number of individuals (the names of individuals).
Else, it is difficult to help without the dataset and the lines of code to read the data and perform the analysis.

FH

--
Vous recevez ce message, car vous êtes abonné au groupe Google Groupes "FactoMineR users".
Pour vous désabonner de ce groupe et ne plus recevoir d'e-mails le concernant, envoyez un e-mail à l'adresse factominer-use...@googlegroups.com.
Cette discussion peut être lue sur le Web à l'adresse https://groups.google.com/d/msgid/factominer-users/c2043d5f-8abe-48ba-be6b-5aa9aa978d19n%40googlegroups.com.

--

	François Husson Department Statistics & Computer science UMR 6625 IRMAR CNRS 65 rue de Saint-Brieuc, CS 84215, 35042 Rennes Cedex Tel: +33 2 23 48 58 86 https://husson.github.io
En 2022, Agrocampus Ouest devient l'Institut Agro Rennes-Angers.

[Robrecht Bollen]

Dear Francis,

that was indeed the problem. I created a supplementary group with a unique code (combining both factors of interest) for each of the individuals. This would then be a "a categorical variable with as many categories as the number of individuals" as you state it. I removed this group and it works fine now, thank you very much.

On another note, I am working with sensory data resulting in groups with variables from descriptive sensory analysis (e.g. chocolate, vanilla...) for groups such a Aroma, Flavor, ... For each sample the amount of times a sensory descriptor is noted results in a frequency table for each of the samples. In essence this is a frequency table for all the different sensory variables. When I indicate in Factoshiny that these variables are 'frequencies' this results in a separate graph for these variables, and are not shown on the correlation circle with the other continuous variables (like quality scores). When I change the sensory groups to 'continuous variables' they are now shown on the correlation circle, making it easier for the reader to interpret the data in relation with the other groups of variables.

My question is: is it possible to show the frequency variables on the correlation circle? And from a statistical point of view would it be incorrect to use them (sensory variables) as continuous variables and show them on the correlation circle?

Thank you and your team for the great program.

Kind regards,

Robrecht Bollen

Op wo 5 okt. 2022 om 16:45 schreef Francois Husson <francoi...@agrocampus-ouest.fr>:

Vous recevez ce message, car vous êtes abonné à un sujet dans le groupe Google Groupes "FactoMineR users".
Pour vous désabonner de ce sujet, visitez le site https://groups.google.com/d/topic/factominer-users/WHfB9AglacE/unsubscribe.
Pour vous désabonner de ce groupe et de tous ses sujets, envoyez un e-mail à l'adresse factominer-use...@googlegroups.com.
Cette discussion peut être lue sur le Web à l'adresse https://groups.google.com/d/msgid/factominer-users/ab7c0907-2daf-8a21-decd-7cd4d82dc709%40agrocampus-ouest.fr.

--

[Francois Husson]

Hi,

If you consider a table crossing the catgories of 2 categorical variables (sae table as the nes used to perform a chi-squared test), it is better to perform correspondence analysis. And yes, in that case, rows and columns are on a same graph. You can see the courses on PCA, CA and the other methods here: https://husson.github.io/MOOC.html#AnaDoGB
So the frequencies are not drawn on the correlation circle.
Sensory variables can be considered as continuous variables if they orrespond to score for instance, but not if, for instance, it corresponds to the number of times panellists  cited a sensory desciptor to describe a product.
Best
FH

Cette discussion peut être lue sur le Web à l'adresse https://groups.google.com/d/msgid/factominer-users/CAEDsBztsMzZr5caa_1HDtVaqchAO5dfZTWxHzwbwkLjWZMPfRQ%40mail.gmail.com.

[Robrecht Bollen]

thank you Francois, have a nice evening.

Op wo 5 okt. 2022 om 20:57 schreef Francois Husson <francoi...@agrocampus-ouest.fr>:

Cette discussion peut être lue sur le Web à l'adresse https://groups.google.com/d/msgid/factominer-users/2ac251cd-9ec9-2ffe-80e2-5bde81e06173%40agrocampus-ouest.fr.

--

[Francois Husson]

Dear Robrecht,

You want to see the link between a qualitative variable and several continuous variables. This objective can be attend by a model such as a logistic model for instance.
With principal component methods such as PCA, the objective is to have a multidimensional approach to better visualize the individuals, their resemblance and differences, and to see the relationships between variables.

So I think you can use MFA or PCA to see your data set, and then you should use a model such as a logistic one.
FH

Le 07/04/2023 à 17:31, Robrecht Bollen a écrit :

Dear François,

I have a follow-up question concerning sensory variables in MFA. In the Wine example of the MFA video your colleague uses the "Origin" of the wine as a supplementary variable (which I understand). But what if you want to use this variable as an active variable?

In my research I combine descriptive sensory notes (frequency table) with quality scores of coffees that have been processed differently (e.g. fermented, non-fermented, ...). Using MFA, I want to assess which variables correlate the most with the Processing method (qualitative variable). When I run the analysis with Processing variable as active I get a significant correlation of groups with the dimensions (Dim1, Dim2,...) but when I set the Processing variable to supplementary the automatic description of the axis is not correlated with this variable.

I find no clear explanation of when to choose if a variable should be active or supplementary. If my main objective is to study the relationship of the processing variable with descriptive and qualitative data I just set this variable to active? 

Thank you very much for your time,

Robrecht

Op woensdag 5 oktober 2022 om 20:57:22 UTC+2 schreef François Husson:

Cette discussion peut être lue sur le Web à l'adresse https://groups.google.com/d/msgid/factominer-users/efeea041-9f12-420b-8c94-dad3738fc05cn%40googlegroups.com.