[Rsiena-help] Possible regression with sienaTimeTest on 2-mode networks

[Ruth M. Ripley]

Dear Guido,

The latest version on r-forge is revision 129, not 125. But I think 125 
should be mostly OK.

There was a major revision of sienaTimeTest code and the new code 
(125 onwards) will not work with sienaFit objects which were created with 
earlier versions. I suspect your object was created with an earlier 
version as it is missing an attribute used in sienaTimeTest. If you do not 
want to re-run, the following should make sienaTimetest work (although 
there may be other incompatibilities):

attr(twomode.ans$f, "periodNos") <- c(1,2)

I will try to remember to make it clear in the changelog in the manual 
when upgrades will be incompatible with existing objects.

Regards,

Ruth
--
Ruth M. Ripley,                         Email:ruth at stats.ox.ac.uk
Dept. of Statistics,                    http://www.stats.ox.ac.uk/~ruth/
University of Oxford,                   Tel:   01865 282851
1 South Parks Road, Oxford OX1 3TG, UK  Fax:   01865 272595


On Thu, 13 Jan 2011, Guido Conaldi wrote:

> Dear Ruth,
>
> Here you have the data and the output of sessionInfo that refers to the
> sienaTimeTest failure.
>
> Also, It was me who asked about notes, and comment() is actually what I
> was looking for. Thank you.
>
> Regards,
> Guido
>
>
> On Thu, 2011-01-13 at 00:41 +0000, Ruth M. Ripley wrote:
>> Dear Guido,
>>
>> I have tried to fix sienaTimeTest etc. It seems I have not succeeded,
>> please send me your data and I will investigate. Please include
>> sessionInfo().
>>
>> BTW, was it you who asked about adding notes to fits? You could try
>> comment(). It is not very nice, but it might suffice.
>>
>> Regards,
>>
>> Ruth
>>
>> On Wed, 12 Jan 2011, Guido Conaldi wrote:
>>
>>> Dear All,
>>>
>>> I was checking whether the latest versions of RSiena/Test had solved the
>>> problem we encountered when using includeTimeDummy() on sienaEffects
>>> objects with 2-mode networks. I found instead a possible regression in
>>> sienaTimeTest().
>>>
>>>
>>> -1-
>>> This is what I get with the latest RSiena/Test and a sienaFit object
>>> with 3 time periods and one two-mode network (and I tried many different
>>> ones):
>>>
>>>
>>>> sienaTimeTest(a.compAll.mainEff)
>>> Error in sienaTimeTest(a.compAll.mainEff) :
>>> You must have at least three time periods to test for non-heterogeneity
>>> across time.
>>>
>>>
>>>> a.compAll.mainEff
>>> Estimates, standard errors and t-statistics for convergence
>>>
>>>                                                  Estimate   Standard   t statistic
>>>                                                              Error
>>>
>>> Rate parameters:
>>>  0.1      Rate parameter period 1                 3.5386  ( 0.2790   )
>>>  0.2      Rate parameter period 2                 5.1183  ( 0.4053   )
>>>
>>> ...
>>>
>>>
>>> With RSiena 1.0.11.120 RSienaTest 1.0.12.123 I get the expected output.
>>>
>>>
>>> -2-
>>> This is the error I get with all versions of RSiena when trying to
>>> include a timedummy in a sienaEffect object with a two-mode network:
>>>
>>>
>>>>
>>> epi.effects[[7]]<-includeTimeDummy(epi.effects[[7]],cycle4,timeDummy="2")
>>>
>>> ** caught segfault ***
>>> address 0x10, cause 'memory not mapped'
>>>
>>> Traceback:
>>> 1: .Call("ChangingCovariates", PACKAGE = pkgname, pData, lapply(f,
>>> unction(x) x$vCovars))
>>> 2: initializeFRAN(z, x, data, effects, prevAns, initC, profileData =
>>> profileData, returnDeps = returnDeps)
>>> 3: x$FRAN(z, x, INIT = TRUE, ...)
>>> 4: robmon(z, x, useCluster, nbrNodes, initC, clusterString,
>>> clusterIter, ...)
>>> 5: siena07(epi.model[[i]], data = epi.data[[i]], effects =
>>> epi.effects[[i]], returnDeps = FALSE, batch = TRUE, verbose = TRUE,
>>> nbrNodes = 3, useCluster = TRUE, initC = TRUE)
>>>
>>> I can provide my objects in case you need them.
>>>
>>> I hope you can help me,
>>> Thanks in advance and a Happy new year to all of you!
>>>
>>> Guido
>>>
>>> _______________________________________________
>>> Rsiena-help mailing list
>>> Rsiena-help at lists.r-forge.r-project.org
>>> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rsiena-help
>>>
>
>