Re: Récapitulatif destiné à factominer-users@googlegroups.com - 1 mise à jour dans 1 sujet

10 views
Skip to first unread message

Kris Lockyear

unread,
Jul 16, 2021, 6:20:14 AM7/16/21
to factomin...@googlegroups.com

K.

From: factomin...@googlegroups.com <factomin...@googlegroups.com>
Sent: 16 July 2021 11:11
To: Destinataires du récapitulatif <factomin...@googlegroups.com>
Subject: Récapitulatif destiné à factomin...@googlegroups.com - 1 mise à jour dans 1 sujet
 
kalongo hamusonde <kalo...@gmail.com>: Jul 16 12:51AM -0700

Hi all,
 
I am having troubles understanding what the two values "Cla/Mod and
"Mod/Cla mean.
I read somewhere that the Cla/mod corresponds to the within cluster
variation. If that's the case, then why are the variables with high values
of Cla/Mod ranked first in clusters. Does this mean that the variables
with the high Cla/Mod values are largely dispersed in that cluster hence
should not be considered ?
Ce résumé vous a été envoyé, car vous êtes abonné aux actualités de ce groupe. Vous pouvez modifier vos paramètres sur la page d'abonnement au groupe.
Pour vous désinscrire de ce groupe et ne plus recevoir de messages, envoyez un e-mail à l'adresse factominer-use...@googlegroups.com.

novio...@hotmail.com

unread,
Jul 21, 2021, 5:53:47 PM7/21/21
to factomin...@googlegroups.com
The larger the dataset, the lower the percentage variance explained tends to be, which makes intuitive sense. Additionally, the usual way this is calculated rather under estimates it as argued by Greenacre (e.g. In CA in practice). His ca package provides options for the alternatives he suggests.

Bootstrapping might be one option. See Ringrose's package cabootcrs although the cases will be overwhelming you can at least see the stability of the variables.

K.



-------- Original message --------
From: factomin...@googlegroups.com
Date: Wed, 21 Jul 2021, 11:10
To: Destinataires du récapitulatif <factomin...@googlegroups.com>
Subject: Récapitulatif destiné à factomin...@googlegroups.com - 1 mise à jour dans 1 sujet
Dominique Beling Nkoumba <belingd...@gmail.com>: Jul 20 08:26AM -0700

Hello everyone,
 
I'm trying to perform an MCA on a large sample of individuals (>15 000) and
8 variables with R. When I perform that MCA using Factominer, my first
dimension 1 inertia is 11 % and the second one is 10 %. Then I want to know
how can I assess of the quality of an MCA. Do any of you has some test
which can help to test the quality of my program ?
 
Best regards,
Dominique.
Reply all
Reply to author
Forward
0 new messages