[R] Problem to predict new data with mclust

[Sarah Goslee]

Many people do not want to download and open random attachments.
That's why I suggested dput() with part of your data.

dput(head(yourdata, 20))

is enough for many problems, though it's good to try it out.

Or use one of the datasets built into R.

Sarah


On Tue, Dec 10, 2013 at 3:23 PM, Welma Pereira <welma.pereira at gmail.com> wrote:
> Hi,
>
> Thanks for the hint Sarah. So here is my problem again: (files train_pca is
> in train_pca and test is in test_pca attached)
>
> mclust2_pca <- Mclust(train_pca,G=2, modelNames= c("EII", "VII", "EEI",
> "EVI", "VEI", "VVI"))
> pred<-predict.Mclust(mclust2_pca,test)
> Warning message:
> In cdensVVI(data = c(2.80217508052409, 2.75071740560707, 2.6175058179515,  :
>   cannot compute E-step
>
> Thanks!
>
>
> On 10 December 2013 16:30, Sarah Goslee <sarah.goslee at gmail.com> wrote:
>>
>> Hi,
>>
>> It's unlikely that anyone will be able to see your problem without a
>> reproducible example.
>>
>>
>> http://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example
>>
>> Are you using the mclust package? That would be the first thing we need to
>> know.
>>
>> Then, have you worked through the examples in ?predict.Mclust, and do
>> you understand them?
>>
>> Then, what do your data look like? str() and such are useful, but the
>> best option is to use dput() to create a reproducible subset, or to
>> reproduce your problem using one of the built-in datasets in R, as is
>> done in the above documentation examples.
>>
>> Sarah
>>
>> On Tue, Dec 10, 2013 at 6:41 AM, Welma Pereira <welma.pereira at gmail.com>
>> wrote:
>> > Hi,
>> >
>> > I am trying to use mclust to cluster some data (train_pca_10), I get the
>> > clusters, but when I try to use the model to predict new data (test1) I
>> > get
>> > this error
>> >
>> >
>> >
>> > mClust2 <- Mclust(train_pca_10,G=2)
>> > pred<-predict.Mclust(mClust2,test1)
>> >
>> > Error in if (warn) warning(WARNING) : argument is of length zero
>> >
>> > Can anyone see the problem here?
>> >
>> > Thanks,
>> > Pereira.
>> >