[Rsiena-help] Interaction effect specification for bipartite networks

[marco tonellato]

Dear Ruth,

That was quick! Meanwhile, I've tried running the model with only one core
and it worked as you predicted. I'll try the new code as soon as it'll be
available. Thanks a lot for your help!

Best,

Marco


On Wed, Mar 17, 2010 at 2:26 PM, Ruth M. Ripley <ruth at stats.ox.ac.uk> wrote:

> Dear Marco,
>
> I have now put the new source on R-forge. I hope it will be built
> by tomorrow.
>
> Regards,
>
> Ruth
> --
> Ruth M. Ripley,                         Email:ruth at stats.ox.ac.uk<Email%3Aruth at stats.ox.ac.uk>
> Dept. of Statistics,                    http://www.stats.ox.ac.uk/~ruth/<http://www.stats.ox.ac.uk/%7Eruth/>
> University of Oxford,                   Tel:   01865 282851
> 1 South Parks Road, Oxford OX1 3TG, UK  Fax:   01865 272595
>
>
> On Wed, 17 Mar 2010, Ruth M. Ripley wrote:
>
> > Dear Marco,
> >
> > I have tried a similar model myself which fails... I know what is wrong:
> > user specified interactions will not work with multiple processes. It
> will
> > take me a little while to fix it, and it is late here. I will try
> > tomorrow!
> >
> > Regards,
> >
> > Ruth
> > --
> > Ruth M. Ripley,                         Email:ruth at stats.ox.ac.uk<Email%3Aruth at stats.ox.ac.uk>
> > Dept. of Statistics,                    http://www.stats.ox.ac.uk/~ruth/<http://www.stats.ox.ac.uk/%7Eruth/>
> > University of Oxford,                   Tel:   01865 282851
> > 1 South Parks Road, Oxford OX1 3TG, UK  Fax:   01865 272595
> >
> >
> > On Tue, 16 Mar 2010, marco tonellato wrote:
> >
> > > Dear Ruth,
> > >
> > > now that I changed the specification as 'ego' effect I could start the
> > > model, but R then crashed.
> > >
> > > This is the command I used and the output I got after the crash:
> > >
> > > >
> > >
> myeff<-includeInteraction(myeff,egoX,outActSqrt,interaction1="dev.ten",include=TRUE)
> > >                                         name shortName type
> interaction1
> > > collaboration.objEffects.138.1 collaboration   unspInt eval
> > >                                interaction2 include effect1 effect2
> effect3
> > > collaboration.objEffects.138.1                 TRUE      24      17
>   0
> > >
> > > > ans <- siena07(mymodel, data=mydata, effects=myeff, batch=TRUE,
> > > verbose=TRUE, nbrNodes=3, useCluster=TRUE, initC=TRUE)
> > >
> > > Stochastic approximation algorithm.
> > > Initial value for gain parameter = 0.3464102.
> > > Start of the algorithm.
> > > Target function values are
> > >   1.  904.00000   2.   89.00000   3. 1009.38249   4. 5637.09646   5.
> > > -23.27556
> > >   6.  354.12048   7.   72.53726   8.  221.19411   9.  -16.00520  10.
> > > 2756.24970
> > >  11. 2756.24970
> > >
> > > Start phase 0
> > > theta: -1.55  0.00  0.00  0.00  0.00  0.00  0.00  0.00  0.00  0.00
>  0.00
> > > Current parameter values:
> > > -1.547494  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000
> > > 0.000000
> > >  0.000000  0.000000  0.000000
> > >
> > > Start phase 1
> > > Error in unserialize(node$con) : error reading from connection
> > >
> > >
> > > Best regards,
> > >
> > > Marco
> > >
> > > On Tue, Mar 16, 2010 at 11:28 PM, Ruth M. Ripley <ruth at stats.ox.ac.uk
> >wrote:
> > >
> > > > Dear Marco,
> > > >
> > > > Sorry, I think the information I have coded in the package is
> incorrect. I
> > > > think the egoX should be an 'ego' effect not a 'dyadic': if you
> change
> > > > this in your effects object it should be OK.
> > > >
> > > > I will fix it in the package but it will be a couple of days before
> it is
> > > > available in the binary on R-forge.
> > > >
> > > > Regards,
> > > >
> > > > Ruth
> > > > --
> > > > Ruth M. Ripley,                         Email:ruth at stats.ox.ac.uk<Email%3Aruth at stats.ox.ac.uk>
> <Email%3Aruth at stats.ox.ac.uk <Email%253Aruth at stats.ox.ac.uk>>
> > > > Dept. of Statistics,
> http://www.stats.ox.ac.uk/~ruth/ <http://www.stats.ox.ac.uk/%7Eruth/><
> http://www.stats.ox.ac.uk/%7Eruth/>
> > > > University of Oxford,                   Tel:   01865 282851
> > > > 1 South Parks Road, Oxford OX1 3TG, UK  Fax:   01865 272595
> > > >
> > > >
> > > > On Tue, 16 Mar 2010, marco tonellato wrote:
> > > >
> > > > > Dear All,
> > > > >
> > > > > I'm currently modeling a bipartite network of 83 x 577 nodes with
> RSiena
> > > > > (revision 0.61). I was able to estimate significant and convergent
> > > > > parameters for the endogenous part and the main effects of 5
> attributes
> > > > > (related to either node set).
> > > > >
> > > > > I now have a problem including interaction effects. In particular I
> would
> > > > > like to include the interaction of a constant binary covariate
> (which
> > > > > captures whether or not my nodes of the first mode have tenure)
> with the
> > > > > outdegree activity.
> > > > >
> > > > > Both using the tcl/tk initerface and entering (also with RSienaTest
> > r62):
> > > > >
> > > > >
> > > >
> myeff<-includeInteraction(myeff,egoX,outActSqrt,interaction1="dev.ten",include=TRUE)
> > > > >
> > > > > R returned this error message:
> > > > >
> > > > > invalid interaction specification:"must be at least one ego or both
> > > > dyadic
> > > > > effects".
> > > > >
> > > > > Finally, I also noticed that the "interactionType"  attribute of
> the
> > > > tenure
> > > > > egoX effect is set as dyadic, whereas it is left unspecified for
> all the
> > > > > altX effects related to the other attributes I measured for the
> nodes of
> > > > the
> > > > > second mode.
> > > > >
> > > > > Am I doing something wrong? What should I do to get the interaction
> > > > effects
> > > > > working?
> > > > > Thanks a lot,
> > > > >
> > > > > Marco
> > > > >
> > > >
> > > >
> > >
> >
> >
> >
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.r-forge.r-project.org/pipermail/rsiena-help/attachments/20100317/1c3ea293/attachment.htm