[Traminer-users] The substitution cost matrix is not symmetric

Matthias Studer Matthias.Studer at unige.ch
Mon Apr 18 11:37:45 CEST 2011


Dear Alexandre,

Many thanks for your bug report and I was able to reproduce it. In fact, 
it seems to me that we should ignore the warning in that case.

It happens that the matrix computed by seqsubm is not exactly symmetric 
due to very small to rounding errors. The differences should be 
extremely low (something like 1e-16). You may check that by using the 
following code:

     subm - t(subm)

The resulting matrix should contains only zero or extremely low values.

Hope this helps
Matthias


Le 15.04.2011 17:29, Alexandre Pollien a écrit :
> Hello,
>
> When I use seqsubm, now I get this message:
>
>>  contact.om<- 
> seqdist(contact.seq,method="OM",indel=2,sm=cout,with.missing=FALSE, 
> full.matrix=TRUE)
> [>] 5764 sequences with 9 distinct events/states
> [>] 496 distinct sequences
> [>] min/max sequence length: 1/12
> [>] computing distances using OM metric
> [>] total time: 1.25 secs
> Message d'avis :
> The substitution cost matrix is not symmetric.
>
> cout is computed with seqsubm (method="TRATE")
>
>
> A surprising aspect is that it is not systematic (3 / 4)
>
> I thought of a problem of missing value
>
>
> Another amazing prompt is that seqdef shows this message:
>>  contact.seq <- seqdef(contact, 2:15, states = contact.shortlab,labels 
> = contact.lab)
> [>] found missing values ('NA') in sequence data
>
> ...even when there is no missing in my data! (How can I seethe 
> sequence containing the missing data?)
>
>
> Does anyone have a idea?
>
> Thank you very much
>
> Best regards and good weekend
>
> Alexandre
>
>
>
>
> _______________________________________________
> Traminer-users mailing list
> Traminer-users at lists.r-forge.r-project.org
> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/traminer-users
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/traminer-users/attachments/20110418/e746e1fd/attachment.htm>


More information about the Traminer-users mailing list