*****  To join INSNA, visit  *****

Some answers... this gets a bit technical, it would have been 
appropriate for [log in to unmask] (which I cc) - but I see that 
over here, you may find more advice on the relational events models.

> 1) I have been using siena in the past and I know that if I want to use the
> multigroup option I need to have the same amount of waves for every group.

Actually, no - you can have different numbers of observation moments for 
each team.

>   I also believe that every wave needs to last the same amount of time (e.g. 5
> minutes). Am I correct?

Also here, the answer is no - the model assumes that the change process 
follows the same rules over groups and periods (unless you differentiate 
those by variables), but how long periods are or how many you have is 
not restricted.

> 2) The assumption, of what a tie is, is different in siena and relational event
> modeling. The former assumes it to be a state (enduring) whereas the latter an
> event (discrete). I believe that in the emergency care teams as network ties are
> measured as person A talking to person B, the assumption of a discrete event
> applies (similar to the radio communication example mentioned in Butts (2008).

This is correct.

> The issue is that this study is part of my PhD and as I'm in the Netherlands, the
> dissertation is in the form of a collection of articles. Another study, which is
> included in the dissertation, uses siena. So I have two chapters, which follow each
> other, with different assumptions for what a tie is. Therefore my arguments need
> to be clear and convincing. In short this is what I argue: "When measuring
> information sharing by asking respondents to state 'how frequently they share
> information with person X over the past 3 months', the tie is operationalized as an
> enduring state.

Sounds fine by me - when you aggregate event data over a certain time 
period (here: past 3 months), you get count data, and when you binarise 
those, you have 'state' type data. You should make sure that you 
binarise the network at a threshold that is not too low, and that 
subsequent discrete measures do not overlap in aggregation period - 
otherwise you get strange stability artefacts into your data.

> In this situation, it is assumed that - if everything else remains
> constant - the tie will continue to exists.

Uh, no, at least not by the Siena-modelling approach. It assumes an 
unobserved change process between observations.

> However, when measuring information
> sharing by looking at every instant of interaction, the tie is not anymore enduring,
> but discrete."

I guess you mean, 'instantaneous'/ short-lived, yes.

>   So my basic argument is that I measure the tie differently, and
> therefore have to use a different method. Is this convincing? Do I miss something?

Hm, I would say that you can easily argue that by aggregation over 3 
months, you get at the more abstract concept "there is an open 
information sharing channel", which is of state type, and you can use 
Siena. When you do not aggregate, you model concrete information sharing 
events, and you cannot but have to work with relational event models.

Which, given that you have time-stamped data, may be the statistically 
more efficient analysis method anyway.


_ __ ___ ____ ___ __ _ __ ___ ____ ___ __ _

Christian Steglich, University of Groningen
Faculty of Behavioural and Social Sciences
Grote Rozenstraat 31, NL-9712 TG Groningen
_ __ ___ ____ ___ __ _ __ ___ ____ ___ __ _

SOCNET is a service of INSNA, the professional association for social
network researchers ( To unsubscribe, send
an email message to [log in to unmask] containing the line
UNSUBSCRIBE SOCNET in the body of the message.