Meaning of and countermeasures for "Stopped after xxiterations" when using imputePCA()

[Chris]

Hi,

I have a dataset with 4680 rows and 146 columns (see attached data file).

Running PCA() on it I get the following Eigenvalues:

My problem is that there are a lot of missing values in the dataset (so these were replaced by the column mean in the PCA() function by default).

Due to the missing values I want to run imputePCA() on the dataset first. Unfortunately it takes too long to estimate the required number of components with estim_ncpPCA (cancelled it after hours). 

Running imputePCA() with these parameters
ncp <- 20
nb.init <- 3
maxiter <- 1000

results in a warning "Stopped after  1000  iterations". What does this mean? Do i need to use a lower number of ncp? Or a higher number of maxiter? Is there any way of getting the "right" number of components for the imputation besides the imputePCA() function?

Thanks a lot for your help.

[josse]

Hi Chris,

Selecting the number of dimensions with missing values is a rough task,
However imputing with 20 dimensions is too much,
I would not go after 5! and often with 2, you get some accurate results.

With estim_ncpPCA, you can play with the options using Kfold cv but it may take time.
So my practical advice would be to use 4 dimensions, you could then assess the stability of the results for instance trying 3 or 5 and you could have a look also at the confidence areas for instance using multiple imputation with MIPCA,

All the best,
Julie.

Le 13/11/14 11:22, Chris a écrit :

--
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.
Pour obtenir davantage d'options, consultez la page https://groups.google.com/d/optout.