SOCNET@LISTS.UFL.EDU

View:

 Message: [ First | Previous | Next | Last ] By Topic: [ First | Previous | Next | Last ] By Author: [ First | Previous | Next | Last ] Font: Proportional Font

Subject:

Re: Structural data based on actor attributes

From:

Date:

Mon, 14 Oct 2002 10:41:46 -0400

Content-Type:

text/plain

Parts/Attachments:

 text/plain (133 lines)
 ```***** To join INSNA, visit http://www.sfu.ca/~insna/ ***** Carl, I should add that I've not read Taylor and don't know how he actually interprets his measure. Based on your note I have assumed it is interpreted as a network tie between cities. It's worth pointing out that products like Taylor's are often used to measure similarity. For example, the Hubert approach to QAP similarity between matrices is based on summing products. The pearson correlation coefficient can be defined as the average of products of standardized variables. So if you wanted to know how similar two cities were in terms of their involvement with TNCs, you could take the absolute value of the difference of their involvements, ABS(O(i)-O(j)), which is a dissimilarity measure. Or, in some cases, you could also take their product O(i)*O(j), which is a crude similarity measure. steve. ----- Original Message ----- From: "Steve Borgatti" <[log in to unmask]> To: <[log in to unmask]> Sent: Monday, October 14, 2002 10:02 AM Subject: Re: Structural data based on actor attributes > ***** To join INSNA, visit http://www.sfu.ca/~insna/ ***** > > Carl, > > In general, I would say this is not a good idea. There are exceptions, but > even then interpretation is tricky since no actual ties have been observed > between the actors. Usually, you must assume that a tie of this type implies > the potential for a real tie. So, for example, if we have a large group of > people that are working together in small teams, and we have a categorical > attribute TEAM which identifies the team for each person, then we can > construct a dyadic variable called SAMETEAM in which ST(i,j) = 1 if T(i) = > T(j) and 0 otherwise. This variable has some merit, particularly if used, > for example, as a control in a dyadic regression. It can be interpreted as > indicating a potential for communication/acquaintanceship between i and j. > Note that if the teams get large (suppose they are prisons or organizations) > then the argument that a sameorg tie implies a social tie gets weak. I think > that the Taylor variable has much less to recommend it than the sameteam or > samegender or samecountryclub variable. > > NOte also that "networks" constructed from node attributes have > characteristics that are artifactual. The sameteam variable is an > equivalence relation (it is reflexive, symmetric, transitive). The Taylor > variable O(i)*O(j) will have a core/periphery structure (in the ordinary > sens)e that is based on number of offices if the variance of O is high > enough. So that if city i and city j both have lots and lots of offices, > there will be strong "tie" between them (and both will be in the core). If > city i has many offices but city j does not, there will be a middling tie, > and both have few offices, there will be weak tie (they belong to the > periphery). So you would not want to analyze the structure of this matrix > and then "discover" that it has a core/periphery structure no doubt created > by imperialist capitalist nations who control the tncs, etc. > > steve. > > ----- Original Message ----- > From: "Carl Nordlund" <[log in to unmask]> > To: <[log in to unmask]> > Sent: Monday, October 14, 2002 7:08 AM > Subject: Structural data based on actor attributes > > > > ***** To join INSNA, visit http://www.sfu.ca/~insna/ ***** > > > > Dear all, > > > > When studying the 'World City Network', Peter Taylor has specified a > > technique for obtaining structural data which is based on actor attributes > > (Research bulletin 23, GaWC, also published in Geographical Analysis > 33(2), > > 2001). He uses a formula like this one: > > > > C(i,j) = O(i) * O(j) > > > > ...where C(i,j) is the total connectivity between actor i and j and where > > O(i) and O(j) are total number of TNC offices of actor i respectively j. > > > > In short, what is done here is an estimation of the structural value > between > > dyads based on internal attributes of the actors. If city i has 2 offices > > and city j has 3 offices, the structural connectivity value between i and > j > > is set to 6, thus implying the total number of links between these two > > actors. > > > > Is this a common way of fetching structural data, i.e to use actor > > attributes like this? Intuitively, what regards Taylor's study, it feels > > like a reasonable good approximation of the structural data but when it is > > only based on actor-internal attributes, I also get the feeling that it > > isn't following a very 'networkish' style! > > > > Yours, > > Carl > > ----- > > Carl Nordlund, BA, PhD student > > carl.nordlund(at)humecol.lu.se > > Human Ecology Division > > www.humecol.lu.se > > > > _____________________________________________________________________ > > SOCNET is a service of INSNA, the professional association for social > > network researchers (http://www.sfu.ca/~insna/). To unsubscribe, send > > an email message to [log in to unmask] containing the line > > UNSUBSCRIBE SOCNET in the body of the message. > > > > _____________________________________________________________________ > SOCNET is a service of INSNA, the professional association for social > network researchers (http://www.sfu.ca/~insna/). To unsubscribe, send > an email message to [log in to unmask] containing the line > UNSUBSCRIBE SOCNET in the body of the message. > _____________________________________________________________________ SOCNET is a service of INSNA, the professional association for social network researchers (http://www.sfu.ca/~insna/). To unsubscribe, send an email message to [log in to unmask] containing the line UNSUBSCRIBE SOCNET in the body of the message.```