imputeFAMD and imputeMCA errors for missing data

685 views
Skip to first unread message

kenc...@tcd.ie

unread,
Jun 19, 2015, 1:07:32 PM6/19/15
to factomin...@googlegroups.com
Hi everyone!

When using a dataset with missing values, after imputing the dataset the hcpc desc.var call shows additional categories for the values which were originally missing (as other users have mentioned). As this seems to be biasing the clustering (these categories are all strongly significantly linked to the cluster descriptions), I have made sure to read the missing data values into R as "na" using "na.strings" in read.csv. Now when I try to run the analysis, I get the error: 

"Error in if (any(MM[[g]] < 0)) stop(paste("The algorithm fails to converge. Choose a number of components (ncp) less or equal than ",  : 
  missing value where TRUE/FALSE needed"

 If anyone has an idea of how to fix this I'd be really grateful! 

Many thanks,
Charlotte

François Husson

unread,
Jun 25, 2015, 11:54:52 AM6/25/15
to factomin...@googlegroups.com, kenc...@tcd.ie
Missing values do not biased the clustering because the clustering is performed with the imputed values.
But when you characterize the clusters, the missing values are considered. If the category corresponding to missing values characterizes a cluster, it is an interesting information : the individuals of this cluster are specific and have more missing values than the individuals of the other clusters.

FH

Charlotte Kenchington

unread,
Jun 25, 2015, 1:38:25 PM6/25/15
to factomin...@googlegroups.com

Hi François,
That is very helpful and interesting.
Do you have any insight into what is causing the MM [[g]] <0 error?
Thanks,
Charlotte

--
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/0foiGBch6Nw/unsubscribe.
Pour vous désabonner de ce groupe et de tous ses sujets, envoyez un e-mail à l'adresse factominer-use...@googlegroups.com.
Pour obtenir davantage d'options, consultez la page https://groups.google.com/d/optout.

FaV1

unread,
Mar 8, 2016, 9:14:23 PM3/8/16
to FactoMineR users
I have the same problem, where you guys able to figure out what the problem is?

E M

unread,
Oct 5, 2020, 1:51:20 PM10/5/20
to FactoMineR users
Yes, what would the solution be then?
Reply all
Reply to author
Forward
0 new messages