- Hello,

I have questions on how to analyze and interpret the indicator

network density.

My data consists of two tables: people and relations. The relations

table links two people at a time, however it is not necessary, that

the second person is in the people table. I have different kinds of

relations, for example "plays with" and "is taught by" (the context

is a gaming community :-)).

Now I would like to interpret the network density. I have:

- Number of signed-up players: 119

- Maximum number of possible relations (119*(119-1)/2): 7021

- Number of relations between signed-up players: 164

- Network density (164/7021): 0.0233584959

I have used the following assumptions:

- neglect direction of relations (treat links as undirected)

- neglect kind of relation (if there are more than one relation

between two people count just one)

- only count links within the population given by table people

Now my questions:

- Are my assumptions reasonable?

- How can I interpret the network density? I have no case study

for comparison.

- What about the behavior of network density in case of

incomplete data? I guess I can assume, that the people who

signed up submitted correct and mostly complete data. But how

should I deal with people who did not signed up? The community

consists of nearly 2000 players. Is it reasonable to generalize

the indicator network density for the whole community?

I guess these questions can easily be conferred to e.g. business

settings, so that they are interesting for you anyhow.

Best regards,

Steffen Mazanek

http://informatik-praxis.blogspot.com - Giancarlo, thank you for your elaborated response. I have to think

about it and will follow up soon.

Best regards,

Steffen

--- In ona-prac@yahoogroups.com, "Giancarlo Oriani"

<giancarlo.oriani@...> wrote:>

interesting of them being probably density, cohesion (average of

> Dear Steffen,

> your question is very interesting.

> We can use different kinds of overall network indicators, the most

distances between any pairs) and E/I.> Each of them cannot be evaluated simply.

about the maximum possible value" (Scott, 2000). But we can have this

> Let's discuss just density.

> First of all, for valued relationhips, we need "some assumptions

assumption, for we are who define the response scale.> Second, being people able to manage a limited numbers of ties, we

expect density to decrease as network size is increasing. Generally,

network density decreases if actor degrees remain unchanged.> Consequently, it is very difficult to compare density indicators of

different network. But we should try. Cross reminds that "when you

interpret network density you must either relate groups of similar

size or determine an ideal network pattern depending on the objectives

of the group". I believe that also the latter one depends on

information about similar networks.> If you have a look at the case studies published on Cross's web

site, you can see that the density target, based on his "Network

Roundtable high performer benchmark database" is generally somewhere

between 10 and 20. Unfortunately, you cannot use it, as he is not

talking about gaming community.> Ciao

> Giancarlo

>

>

>

>

>

>

>

> ----- Original Message -----

> From: steffenmazanek

> To: ona-prac@yahoogroups.com

> Sent: Saturday, August 12, 2006 5:32 PM

> Subject: [ona-prac] Re: network density

>

>

> Dear ona-practitioners,

>

> I would like to restate and extend my question on network density.

> What kinds of overall network indicators do you use and how do you

> evaluate them? Do you know resources where I can find and study

> examples for comparison? I would be glad if you could share your

> experiences with me.

>

> Best regards,

>

> Steffen Mazanek

>

> http://informatik-praxis.blogspot.com

>

> --- In ona-prac@yahoogroups.com, "steffenmazanek" <smazanek@> wrote:

> >

> > Hello,

> >

> > I have questions on how to analyze and interpret the indicator

> > network density.

> >

> > My data consists of two tables: people and relations. The relations

> > table links two people at a time, however it is not necessary, that

> > the second person is in the people table. I have different kinds of

> > relations, for example "plays with" and "is taught by" (the context

> > is a gaming community :-)).

> >

> > Now I would like to interpret the network density. I have:

> > - Number of signed-up players: 119

> > - Maximum number of possible relations (119*(119-1)/2): 7021

> > - Number of relations between signed-up players: 164

> > - Network density (164/7021): 0.0233584959

> >

> > I have used the following assumptions:

> > - neglect direction of relations (treat links as undirected)

> > - neglect kind of relation (if there are more than one relation

> > between two people count just one)

> > - only count links within the population given by table people

> >

> > Now my questions:

> > - Are my assumptions reasonable?

> > - How can I interpret the network density? I have no case study

> > for comparison.

> > - What about the behavior of network density in case of

> > incomplete data? I guess I can assume, that the people who

> > signed up submitted correct and mostly complete data. But how

> > should I deal with people who did not signed up? The community

> > consists of nearly 2000 players. Is it reasonable to generalize

> > the indicator network density for the whole community?

> >

> > I guess these questions can easily be conferred to e.g. business

> > settings, so that they are interesting for you anyhow.

> >

> > Best regards,

> >

> > Steffen Mazanek

> >

> > http://informatik-praxis.blogspot.com

> >

>