Print

Print


*****  To join INSNA, visit http://www.insna.org  *****

Dear all,

Two further cents to this discussion:

A. Although I agree with Johan's general point, I think that in practice 
the distinction is smaller than he suggests. This must be linked to 
Garry's remark that all that can be concluded from this type of 
cross-sectional data is association between attribute and network 
structure, not anything about the causal directionality which is 
mentioned in Jorge's question. Let us call the network X and the 
attribute Y. The null hypothesis of absence of association, i.e., 
independence between X and Y, can be tested by using X as an independent 
variable and Y as the dependent variable, but also by using Y as the 
independent and X as the dependent variable. This is like in regression 
and correlation analysis, where the correlation between X and Y is zero 
if and only if the regression coefficient of X and Y is zero, which 
again is the case if and only if the regression coefficient of Y on X is 
zero.
Concluding, I agree with Johan's remark that "the method should mirror 
the substantive issue". But I would disagree with the implied suggestion 
(which however is not explicit in Johan's message) that absence of 
association between network and attribute cannot properly be tested by 
testing homophily using an ERG model.

B. Filip asks the rhetorical question, "why not use a regression model 
with these opinions as dependent variables and network structure as 
independent?". I would like to answer as follows. Given the above, the 
use of such a regression model is a reasonable way for testing the null 
hypothesis of absence of association between opinions and network. On 
the other hand, if there is such an association (which we conclude if 
the result is significant), then the association between network and 
opinions will lead to network-related dependencies between the opinions; 
it is quite unlikely that these are totally represented by the 
network-related variables used as "independent" variables in the 
regression. This implies that the residuals in the regression will be 
dependent, and an OLS regression model then is likely to be misspecified 
and its results unreliable.
This misspecification holds under the alternative hypothesis (of 
association between attributes and network), not under the null 
hypothesis. Therefore, as a null hypothesis test, this approach is OK. 
But we cannot use this approach e.g.  for obtaining reliable conclusions 
about the type of dependence between attributes and network, nor about 
the association between the opinions and other attributes used as 
control variables while controlling for network dependencies.
The network autocorrelation models also mentioned by Filip do give an 
explicit representation of the network dependencies. Although the 
question of specification and misspecification always remains, these at 
least provide a theoretically consistent model of the network-attribute 
association which also can be employed if there is indeed association 
between the network and the attribute.

Cheerio,

Tom


On 03/05/2011 10:03, Johan Koskinen wrote:
> ---------------------- Information from the mail header -----------------------
> Sender:       Social Networks Discussion Forum<[log in to unmask]>
> Poster:       Johan Koskinen<[log in to unmask]>
> Subject:      Re: ergm question
> -------------------------------------------------------------------------------
>
> --_000_E3B21BD4B98F0E49905EF5CB1CE78695C92FMBXP09dsmanacuk_
> Content-Type: text/plain; charset="Windows-1252"
> Content-Transfer-Encoding: quoted-printable
>
> *****  To join INSNA, visit http://www.insna.org  *****
>
> Jorge,
>
> To further add to previous comments, it may be good to keep in mind two thi=
> ngs. Firstly, the method should mirror the substantive issue. If you are in=
> terested in explaining nodal attributes you should use a model for nodal at=
> tributes and not a model, like ERGM, for tie-variables. This is much in the=
>   same way that if you are interested in income differences for men and wome=
> n you would regress income on gender and not the other way around (i.e. reg=
> ressing gender on income). Secondly, if you want to explain attributes, you=
>   model attributes of, say, n individuals. If you want to explain tie-variab=
> les, you are modelling (in the undirected case) n times n-1 tie-variables. =
> Consequently, if you draw inference about your covariates based on an ERG m=
> odel for tie-variables, in a sense you are using your actor attributes n-1 =
> times. You may for example have 25 homophilous tied pairs of actors for a b=
> inary attribute (for or against) but only 10 individuals in total. An ERGM =
> would treat the homophilous (and heterophilous, naturally) ties as observat=
> ions.
>
> In addition to the techniques mentioned previously (QAP and network effects=
> /autocorrelation models), an attribute model analogous to the ERGM for tie-=
> variables is derived in Robins, G., Pattison, P.,&  Elliot, P. (2001). Netw=
> ork models for social influence processes. Psychometrika, 66 , 161=96190. (=
> A related model is described and elaborated in a chapter of a book on ERGM =
> that is forthcoming)
>
> j.
>
> The email [log in to unmask] WILL EXPIRE on 30 April 2011
>
> ************************************************************
> Social Statistics Discipline Area
> Humanities Bridgeford Street
> University of Manchester
> MANCHESTER M13 9PL
> +44 (0)161 30 66 953 (ManU, Kantorovich Building)
> johan.koskinen[at]manchester.ac.uk
> ________________________________
> From: Social Networks Discussion Forum [[log in to unmask]] on behalf of =
> Garry Robins [[log in to unmask]]
> Sent: 03 May 2011 04:57
> To: [log in to unmask]
> Subject: [SOCNET] FW: ergm question
>
> ***** To join INSNA, visit http://www.insna.org *****
> Jorge,
>
> There is nothing wrong with what you are doing. But just as with inferring =
> causality from a standard regression, you can=92t simply infer that the =93=
> pattern of social interactions influences the type of opinion individuals h=
> ave=94.  What you can say, though, is that there is an association between =
> the attribute (opinions) and network structure. This is all we can do with =
> cross-sectional data. It may or may not be enough for your research purpose=
> s. Of course, if you are specifically investigating influence hypotheses, t=
> hen it makes sense to use an influence model rather than a selection model =
> =96 and Tom gives some good suggestions. I=92m opposed to using simple regr=
> essions here, Tom is right to point to the issue of dependency.
>
> BTW there is an ERGM-based influence-type model in the pnet suite of progra=
> ms, akin to a logistic regression regression but taking into account the ne=
> twork dependencies. It needs a binary attribute, however.
>
> Good luck with the research
>
> Garry
>
>
>
>
> Dr Garry Robins
>
> Associate Professor and Reader
> Psychological Sciences
> University of Melbourne
> Victoria 3010
> Australia
>
> Tel: 61 3 8344 4454
> Fax: 61 3 9347 6618
> Web: http://www.psych.unimelb.edu.au/people/staff/RobinsG.html
> Melnet website: http://www.sna.unimelb.edu.au/
>
>
>
>
>
> ________________________________
> From: Social Networks Discussion Forum [mailto:[log in to unmask]] On Beh=
> alf Of Thomas Valente
> Sent: Tuesday, 3 May 2011 4:32 AM
> To: [log in to unmask]
> Subject: Re: ergm question
>
> All
> I think Filip=92s suggestions are quite helpful and at the same time Jorge=
> =92s frustrations understandable.  The problem is some people will be quite=
>   concerned about network dependencies and the challenges with specifying ne=
> twork effects whereas others will be content to focus on behaviors and poli=
> cy interpretations of these types of influences.  (Somewhat like the balanc=
> e between Type I and Type II error.)  And the researcher wishes to =93do th=
> e right thing.=94  We/I have opted to do both, node level regressions using=
>   somewhat standard multi-level (or hierarchical) regression techniques and,=
>   when possible, build models using STATNET and SIENA.  So far the results a=
> re fairly consistent with different approaches yielding different but not c=
> ontradictory insights.  An example is our paper on the adoption/diffusion o=
> f the FCTC:
> Wipfli, H., Fujimoto, K.,&  Valente, T. W. (2010).  Global tobacco control =
> diffusion: The case of the framework convention on tobacco control. America=
> n Journal of Public Health, 100, 1260-1266.
> So there is no right answer and no single approach that covers all situatio=
> ns.  The researcher will need to pursue at least 2 and sometimes multiple m=
> odeling approaches to be confident the findings are consistent across multi=
> ple specifications.
> -Tom
>
> Thomas W. Valente, PhD
> Current:  =C9cole des hautes =E9tudes en sant=E9 publique (Rennes/Paris, Fr=
> ance)
> Usual:
> Director, Master of Public Health Program      http://www.usc.edu/medicine/=
> mph/
> Professor, Department of Preventive Medicine
> Keck School of Medicine
> University of Southern California
> 1000 S. Fremont Ave., #8
> Building A Room 5133
> Alhambra CA 91803
> phone: (626) 457-4139; cell: (626) 429-4123
> email: [log in to unmask]
>
> Social Networks and Health: Models, Methods, and Applications:
>         http://www.oup.com/us (promo code: 28569)
> Evaluating Health Promotion Programs: www.oup-usa.org/
> Network Models of the Diffusion of Innovations: www.hamptonpress.com
> My personal webpage: http://www-hsc.usc.edu/~tvalente/
> The Empirical Networks Project: http://ipr1.hsc.usc.edu/networks/
> You Tube video on Diffusion of Innovations: http://www.youtube.com/watch?v=
> =3DZG9dAIBd4xQ
>
> From: Social Networks Discussion Forum [mailto:[log in to unmask]] On Beh=
> alf Of F.Agneessens
> Sent: Monday, May 02, 2011 7:58 PM
> To: [log in to unmask]
> Subject: Re: ergm question
>
> ***** To join INSNA, visit http://www.insna.org ***** Dear Jorge,
>
> If you are interested in explaining individual outcome, why not use a regre=
> ssion model with these opinions as dependent variables and network structur=
> e as independent?
> This has been done quite extensively: e.g.,
> - Brass, D.J. 1984. Being in the right place. A structural analysis of indi=
> vidual influence in an organization. Administrative Science Quarterly 29: 5=
> 18-539.
> - Sparrowe, R.T, Liden, R.C., Wayne, S.J., Kraimer, M.L. 2001. Social netwo=
> rks and the performance of individuals and groups. Academy of Management Jo=
> urnal: 316-325.
>
> If you do not want to rely on the significance test of the 'normal' regress=
> ion (which traditionally relies on a random sampling from a population), yo=
> u could even make use of a QAP-like node level regression. For example in U=
> CINET you can do this with the following command: Tools>Testing hypotheses>=
> Node Level>Regression.
>
> However, if you think a person's opinion is influenced by the opinions of h=
> is connections you can use an 'autocorrelation'-type of regression. In that=
>   case, see for example:
> -Doreian, P. 1981. Estimating linear models with spatially distributed data=
> . Sociological Methodology 12:359-388.
> -Doreian, P., K. Teuter, C. Wang. 1984. Network autocorrelation models: Som=
> e Monte Carlo results. Sociological Methods&  Research 13: 155-200.
> -Marsden, P.V., Friedkin, N.E. 1993. Network studies of social influence. S=
> ociological Methods&  Research 22, 127-151.
> -Leenders, R.Th.A.J. 2002. Modelling social influence through network autoc=
> orrelation: Constructing the weight matrix. Social Networks 24: 21-47.
>
> Hope this helps,
> Filip
>
> On 02-05-11, Jorge M Rocha<[log in to unmask]<mailto:[log in to unmask]>=
>> wrote:
> ***** To join INSNA, visit http://www.insna.org *****
> Dear all,
> I have a conceptual question that is probably somewhat na=EFve regarding th=
> e analytic logic behind ERGM=92s (I=92ve been using statnet and its suite o=
> f tools).  If someone in the community could point me in the right directio=
> n literature wise, I would appreciate it.  Now let me try to state my quest=
> ion/doubt as briefly and clearly as I can:
>  From what I understand, the modeling framework of an ERGM treats the observ=
> ed network as the dependent variable and the specified structural configura=
> tions and covariate information about nodes/edges as independent variables =
> whose presence/absence increases the log-likelihood for the model.  In othe=
> r words, what we want to achieve in these models is to gain a better unders=
> tanding of the types of processes that might have gone into generating a ne=
> twork similar to the observed one; hypotheses are about this or that config=
> uration or node/edge attribute and their effect on the pattern of social in=
> teractions represented in the network. If so, how would I go about testing =
> hypotheses concerning the effects of social interactions on nodal attribute=
> s (I=92m particularly thinking of different types of opinions individuals m=
> ight tend to have based on their interactions)?
> What I=92ve done so far is kind of arguing backwards (or at least that is h=
> ow it feels to me =96maybe because I=92m coming from a traditional linear/l=
> ogit regression background, and probably part of my problem is thinking in =
> terms of DV and IV).  If the coefficients are significant for some node cov=
> ariate of my interest in the ERGM estimation, then I=92ve been interpreting=
>   this as evidence that it is not unreasonable to argue that the pattern of =
> social interactions influences the type of opinion individuals have (I know=
>   that if I had longitudinal data this issue of reverse causality would not =
> be so much of an issue, but so far all I have is cross-sectional data).  Ho=
> w wrong am I on this (the backward-arguing)?
> Finally, part of my worry is that I=92ll be sending the research to a consu=
> mer behavior/marketing journal where the familiarity with social network st=
> uff =96let alone ERGM=92s=97is limited.  Any suggestions on how to best exp=
> lain these issues to a non-specialized audience?
> In advance, thanks,
>
> Jorge M Rocha
> EGADE Business School
>
>
>
>
> _____________________________________________________________________ SOCNE=
> T is a service of INSNA, the professional association for social network re=
> searchers (http://www.insna.org). To unsubscribe, send an email message to =
> [log in to unmask]<mailto:[log in to unmask]>  containing the line U=
> NSUBSCRIBE SOCNET in the body of the message.
>
> --
>
> Dr. Filip Agneessens
> Department of Sociology/ICS
> University of Groningen
> Grote Rozenstraat 31
> 9712 TG Groningen
> The Netherlands
> email: [log in to unmask]<mailto:[log in to unmask]>
> http://filipagneessens.wordpress.com/
> _____________________________________________________________________ SOCNE=
> T is a service of INSNA, the professional association for social network re=
> searchers (http://www.insna.org). To unsubscribe, send an email message to =
> [log in to unmask]<mailto:[log in to unmask]>  containing the line U=
> NSUBSCRIBE SOCNET in the body of the message.
> _____________________________________________________________________ SOCNE=
> T is a service of INSNA, the professional association for social network re=
> searchers (http://www.insna.org). To unsubscribe, send an email message to =
> [log in to unmask] containing the line UNSUBSCRIBE SOCNET in the body o=
> f the message. ____________________________________________________________=
> _________ SOCNET is a service of INSNA, the professional association for so=
> cial network researchers (http://www.insna.org). To unsubscribe, send an em=
> ail message to [log in to unmask] containing the line UNSUBSCRIBE SOCNE=
> T in the body of the message.
>
> _____________________________________________________________________
> SOCNET is a service of INSNA, the professional association for social
> network researchers (http://www.insna.org). To unsubscribe, send
> an email message to [log in to unmask] containing the line
> UNSUBSCRIBE SOCNET in the body of the message.
>
> --_000_E3B21BD4B98F0E49905EF5CB1CE78695C92FMBXP09dsmanacuk_
> Content-Type: text/html; charset="Windows-1252"
> Content-Transfer-Encoding: quoted-printable
>
> *****  To join INSNA, visit http://www.insna.org  *****
> <html dir=3D"ltr">
> <head>
> <meta http-equiv=3D"Content-Type" content=3D"text/html; charset=3DWindows-1=
> 252">
> <style id=3D"owaParaStyle" type=3D"text/css"></style>
> </head>
> <body ocsi=3D"0" fpstyle=3D"1">
> <div style=3D"direction: ltr; font-family: Tahoma; color: rgb(0, 0, 0); fon=
> t-size: 13px;">
> <div style=3D"">Jorge,<br>
> <br>
> To further add to previous comments, it may be good to keep in mind two thi=
> ngs. Firstly, the method should mirror the substantive issue. If you are in=
> terested in explaining nodal attributes you should use a model for nodal at=
> tributes and not a model, like ERGM,
>   for tie-variables. This is much in the same way that if you are interested=
>   in income differences for men and women you would regress income on gender=
>   and not the other way around (i.e. regressing gender on income). Secondly,=
>   if you want to explain attributes,
>   you model attributes of, say, n individuals. If you want to explain tie-va=
> riables, you are modelling (in the undirected case) n times n-1 tie-variabl=
> es. Consequently, if you draw inference about your covariates based on an E=
> RG model for tie-variables, in a
>   sense you are using your actor attributes n-1 times. You may for example h=
> ave 25 homophilous tied pairs of actors for a binary attribute (for or agai=
> nst) but only 10 individuals in total. An ERGM would treat the homophilous =
> (and heterophilous, naturally) ties
>   as observations.<br>
> <br>
> In addition to the techniques mentioned previously (QAP and network effects=
> /autocorrelation models), an attribute model analogous to the ERGM for tie-=
> variables is derived in Robins, G., Pattison, P.,&amp; Elliot, P. (2001). =
> Network models for social influence
>   processes. Psychometrika, 66 , 161=96190. (A related model is described an=
> d elaborated in a chapter of a book on ERGM that is forthcoming)<br>
> <br>
> j.<br>
> </div>
> <div><br>
> <div class=3D"BodyFragment"><font size=3D"2"><span style=3D"font-size: 10pt=
> ;">
> <div class=3D"PlainText">The email [log in to unmask] WILL EX=
> PIRE on 30 April 2011<br>
> <br>
> ************************************************************<br>
> Social Statistics Discipline Area<br>
> Humanities Bridgeford Street<br>
> University of Manchester<br>
> MANCHESTER M13 9PL<br>
> &#43;44 (0)161 30 66 953 (ManU, Kantorovich Building)<br>
> johan.koskinen[at]manchester.ac.uk<br>
> </div>
> </span></font></div>
> </div>
> <div style=3D"font-family: Times New Roman; color: rgb(0, 0, 0); font-size:=
>   16px;">
> <hr tabindex=3D"-1">
> <div style=3D"direction: ltr;" id=3D"divRpF539720"><font size=3D"2" color=
> =3D"#000000" face=3D"Tahoma"><b>From:</b>  Social Networks Discussion Forum =
> [[log in to unmask]] on behalf of Garry Robins [[log in to unmask]]<b=
> r>
> <b>Sent:</b>  03 May 2011 04:57<br>
> <b>To:</b>  [log in to unmask]<br>
> <b>Subject:</b>  [SOCNET] FW: ergm question<br>
> </font><br>
> </div>
> <div></div>
> <div>***** To join INSNA, visit http://www.insna.org *****<style>=0A=
> <!--=0A=
> @font-face=0A=
> 	{font-family:"MS Mincho"}=0A=
> @font-face=0A=
> 	{font-family:"Arial Unicode MS"}=0A=
> @font-face=0A=
> 	{font-family:Tahoma}=0A=
> @font-face=0A=
> 	{font-family:Cambria}=0A=
> @font-face=0A=
> 	{font-family:Calibri}=0A=
> @font-face=0A=
> 	{font-family:"\@MS Mincho"}=0A=
> @font-face=0A=
> 	{font-family:"\@Arial Unicode MS"}=0A=
> p.MsoNormal, li.MsoNormal, div.MsoNormal=0A=
> 	{margin:0cm;=0A=
> 	margin-bottom:.0001pt;=0A=
> 	font-size:12.0pt;=0A=
> 	font-family:"Times New Roman"}=0A=
> a:link, span.MsoHyperlink=0A=
> 	{color:blue;=0A=
> 	text-decoration:underline}=0A=
> a:visited, span.MsoHyperlinkFollowed=0A=
> 	{color:purple;=0A=
> 	text-decoration:underline}=0A=
> p.MsoAutoSig, li.MsoAutoSig, div.MsoAutoSig=0A=
> 	{margin:0cm;=0A=
> 	margin-bottom:.0001pt;=0A=
> 	font-size:12.0pt;=0A=
> 	font-family:"Times New Roman"}=0A=
> p=0A=
> 	{margin-right:0cm;=0A=
> 	margin-left:0cm;=0A=
> 	font-size:12.0pt;=0A=
> 	font-family:"Times New Roman"}=0A=
> span.EmailStyle18=0A=
> 	{font-family:Calibri;=0A=
> 	color:windowtext}=0A=
> span.EmailStyle19=0A=
> 	{font-family:"Times New Roman";=0A=
> 	color:windowtext;=0A=
> 	font-weight:normal;=0A=
> 	font-style:normal;=0A=
> 	text-decoration:none;=0A=
> 	text-decoration:none}=0A=
> @page Section1=0A=
> 	{margin:72.0pt 72.0pt 72.0pt 72.0pt}=0A=
> -->=0A=
> BODY {direction: ltr;font-family: Tahoma;color: #000000;font-size: 13px;}P =
> {margin-top:0;margin-bottom:0;}</style>
> <div class=3D"Section1">
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">Jorge,</span></font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">&nbsp;</span></font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">There is nothing wrong with what you are doing. But =
> just as with inferring causality from a standard regression, you can=92t si=
> mply infer that the =93pattern of social interactions
>   influences the type of opinion individuals have=94.<span style=3D"">&nbsp=
> ;</span>What you can say, though, is that there is an association between t=
> he attribute (opinions) and network structure. This is all we can do with c=
> ross-sectional data. It may or may not be
>   enough for your research purposes. Of course, if you are specifically inve=
> stigating influence hypotheses, then it makes sense to use an influence mod=
> el rather than a selection model =96 and Tom gives some good suggestions. I=
> =92m opposed to using simple regressions
>   here, Tom is right to point to the issue of dependency.</span></font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">&nbsp;</span></font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">BTW there is an ERGM-based influence-type model in t=
> he
> <span class=3D"SpellE">pnet</span>  suite of programs, akin to a logistic re=
> gression
> <span class=3D"SpellE">regression</span>  but taking into account the networ=
> k dependencies. It needs a binary attribute, however.</span></font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">&nbsp;</span></font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">Good luck with the research</span></font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">&nbsp;</span></font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">Garry</span></font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">&nbsp;</span></font></p>
> <div>
> <div>
> <p style=3D"margin-bottom: 0.0001pt;"><i style=3D""><font size=3D"2" face=
> =3D"Times New Roman"><span style=3D"font-size: 10pt; font-style: italic;">&=
> nbsp;</span></font></i></p>
> <p style=3D"margin: 0cm 0cm 0.0001pt;"><i style=3D""><font size=3D"2" face=
> =3D"Times New Roman"><span style=3D"font-size: 10pt; font-style: italic;">D=
> r Garry Robins</span></font></i></p>
> <p class=3D"MsoAutoSig"><i style=3D""><font size=3D"2" face=3D"Times New Ro=
> man"><span style=3D"font-size: 10pt; font-style: italic;">Associate Profess=
> or and Reader<br>
> Psychological Sciences<br>
> University of Melbourne<br>
> Victoria 3010<br>
> Australia<br>
> <br>
> Tel: 61 3 8344 4454<br>
> Fax: 61 3 9347 6618<br>
> Web:<a href=3D"http://www.psych.unimelb.edu.au/people/staff/RobinsG.html" =
> target=3D"_blank">
> http://www.psych.unimelb.edu.au/people/staff/RobinsG.html</a><br>
> Melnet website:<a href=3D"http://www.sna.unimelb.edu.au/" target=3D"_blank=
> ">http://www.sna.unimelb.edu.au/</a><br>
> <br>
> <br>
> <br>
> &nbsp;</span></font></i><i style=3D""><font size=3D"2"><span style=3D"font=
> -size: 10pt; font-style: italic;"></span></font></i></p>
> </div>
> </div>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">&nbsp;</span></font></p>
> <div>
> <div class=3D"MsoNormal" style=3D"text-align: center;" align=3D"center"><fo=
> nt size=3D"3" face=3D"Times New Roman"><span style=3D"font-size: 12pt;">
> <hr tabindex=3D"-1" size=3D"2" align=3D"center" width=3D"100%">
> </span></font></div>
> <p class=3D"MsoNormal"><b><font size=3D"2" face=3D"Tahoma"><span style=3D"f=
> ont-size: 10pt; font-family: Tahoma; font-weight: bold;">From:</span></font=
>> </b><font size=3D"2" face=3D"Tahoma"><span style=3D"font-size: 10pt; font-=
> family: Tahoma;">  Social Networks Discussion Forum
>   [mailto:[log in to unmask]]<b><span style=3D"font-weight: bold;">On Beh=
> alf Of</span>
> </b>Thomas Valente<br>
> <b><span style=3D"font-weight: bold;">Sent:</span></b>  Tuesday, 3 May 2011 =
> 4:32 AM<br>
> <b><span style=3D"font-weight: bold;">To:</span></b>  [log in to unmask]<b=
> r>
> <b><span style=3D"font-weight: bold;">Subject:</span></b>  Re: ergm question=
> </span></font></p>
> </div>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">&nbsp;</span></font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Cambria"><span style=3D"fon=
> t-size: 12pt; font-family: Cambria;">All</span></font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Cambria"><span style=3D"fon=
> t-size: 12pt; font-family: Cambria;">I think Filip=92s suggestions are quit=
> e helpful and at the same time Jorge=92s frustrations understandable.&nbsp;=
>   The problem is some people will be quite concerned
>   about network dependencies and the challenges with specifying network effe=
> cts whereas others will be content to focus on behaviors and policy interpr=
> etations of these types of influences.&nbsp; (Somewhat like the balance bet=
> ween Type I and Type II error.)&nbsp;And
>   the researcher wishes to =93do the right thing.=94&nbsp; We/I have opted t=
> o do both, node level regressions using somewhat standard multi-level (or h=
> ierarchical) regression techniques and, when possible, build models using S=
> TATNET and SIENA.&nbsp; So far the results are
>   fairly consistent with different approaches yielding different but not con=
> tradictory insights.&nbsp; An example is our paper on the adoption/diffusio=
> n of the FCTC:</span></font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Cambria"><span style=3D"fon=
> t-size: 12pt; font-family: Cambria;">Wipfli, H., Fujimoto, K.,&amp;<b><spa=
> n style=3D"font-weight: bold;">
> </span></b>Valente, T. W. (2010).&nbsp; Global tobacco control diffusion: T=
> he case of the framework convention on tobacco control.
> <u>American Journal of Public Health</u>,<u>100</u>, 1260-1266.</span></fo=
> nt></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Cambria"><span style=3D"fon=
> t-size: 12pt; font-family: Cambria;">So there is no right answer and no sin=
> gle approach that covers all situations.&nbsp; The researcher will need to =
> pursue at least 2 and sometimes multiple modeling
>   approaches to be confident the findings are consistent across multiple spe=
> cifications.</span></font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Cambria"><span style=3D"fon=
> t-size: 12pt; font-family: Cambria;">-Tom</span></font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">&nbsp;</span></font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">Thomas W. Valente, PhD</span></font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">Current:&nbsp; =C9cole des hautes =E9tudes en sant=
> =E9 publique (Rennes/Paris, France)</span></font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">Usual:</span></font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">Director, Master of Public Health Program&nbsp;&nbsp=
> ;&nbsp;&nbsp;&nbsp; http://www.usc.edu/medicine/mph/</span></font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">Professor, Department of Preventive Medicine</span><=
> /font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">Keck</span></font>  School of Medicine</p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">University</span></font>  of Southern California</p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">1000 S. Fremont Ave., #8</span></font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">Building A Room 5133&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&n=
> bsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp=
> ;&nbsp;&nbsp;&nbsp;
> </span></font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">Alhambra</span></font>  CA 91803&nbsp;&nbsp;&nbsp;&nb=
> sp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
> </p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">phone: (626) 457-4139; cell: (626) 429-4123</span></=
> font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">email: [log in to unmask]</span></font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">&nbsp;</span></font></p>
> <p class=3D"MsoNormal"><u><font size=3D"3" face=3D"Times New Roman"><span s=
> tyle=3D"font-size: 12pt;">Social Networks and Health: Models, Methods, and =
> Applications</span></font></u>:</p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; http://www.oup.=
> com/us (promo code: 28569)</span></font></p>
> <p class=3D"MsoNormal"><u><font size=3D"3" face=3D"Times New Roman"><span s=
> tyle=3D"font-size: 12pt;">Evaluating Health Promotion Programs</span></font=
>> </u>: www.oup-usa.org/</p>
> <p class=3D"MsoNormal"><u><font size=3D"3" face=3D"Times New Roman"><span s=
> tyle=3D"font-size: 12pt;">Network Models of the Diffusion of Innovations</s=
> pan></font></u>: www.hamptonpress.com</p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Calibri"><span style=3D"fon=
> t-size: 12pt; font-family: Calibri;">My personal webpage: http://www-hsc.us=
> c.edu/~tvalente/</span></font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Calibri"><span style=3D"fon=
> t-size: 12pt; font-family: Calibri;">The Empirical Networks Project:
> <a href=3D"http://ipr1.hsc.usc.edu/networks/" target=3D"_blank">http://ipr1=
> .hsc.usc.edu/networks/</a></span></font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">You Tube video on Diffusion of Innovations:
> <a href=3D"http://www.youtube.com/watch?v=3DZG9dAIBd4xQ" target=3D"_blank">=
> http://www.youtube.com/watch?v=3DZG9dAIBd4xQ</a></span></font></p>
> <p class=3D"MsoNormal"><font size=3D"2" face=3D"Calibri"><span style=3D"fon=
> t-size: 11pt; font-family: Calibri;">&nbsp;</span></font></p>
> <p class=3D"MsoNormal"><b><font size=3D"2" face=3D"Tahoma"><span style=3D"f=
> ont-size: 10pt; font-family: Tahoma; font-weight: bold;">From:</span></font=
>> </b><font size=3D"2" face=3D"Tahoma"><span style=3D"font-size: 10pt; font-=
> family: Tahoma;">  Social Networks Discussion Forum
>   [mailto:[log in to unmask]]<b><span style=3D"font-weight: bold;">On Beh=
> alf Of</span>
> </b>F.Agneessens<br>
> <b><span style=3D"font-weight: bold;">Sent:</span></b>  Monday, May 02, 2011=
>   7:58 PM<br>
> <b><span style=3D"font-weight: bold;">To:</span></b>  [log in to unmask]<b=
> r>
> <b><span style=3D"font-weight: bold;">Subject:</span></b>  Re: ergm question=
> </span></font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">&nbsp;</span></font></p>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">***** To join INSNA, visit
> <a href=3D"http://www.insna.org" target=3D"_blank">http://www.insna.org</a>=
>   ***** Dear Jorge,<br>
> <br>
> If you are interested in explaining individual outcome, why not use a regre=
> ssion model with these opinions as dependent variables and network structur=
> e as independent?
> <br>
> This has been done quite extensively: e.g.,<br>
> - Brass, D.J. 1984. Being in the right place. A structural analysis of indi=
> vidual influence in an organization. Administrative Science Quarterly 29: 5=
> 18-539.<br>
> - Sparrowe, R.T, Liden, R.C., Wayne, S.J., Kraimer, M.L. 2001. Social netwo=
> rks and the performance of individuals and groups. Academy of Management Jo=
> urnal: 316-325.<br>
> <br>
> If you do not want to rely on the significance test of the 'normal' regress=
> ion (which traditionally relies on a random sampling from a population), yo=
> u could even make use of a QAP-like node level regression. For example in U=
> CINET you can do this with the following
>   command: Tools&gt;Testing hypotheses&gt;Node Level&gt;Regression.<br>
> <br>
> However, if you think a person's opinion is influenced by the opinions of h=
> is connections you can use an 'autocorrelation'-type of regression. In that=
>   case, see for example:<br>
> -Doreian, P. 1981. Estimating linear models with spatially distributed data=
> . Sociological Methodology 12:359-388.<br>
> -Doreian, P., K. Teuter, C. Wang. 1984. Network autocorrelation models: Som=
> e Monte Carlo results. Sociological Methods&amp; Research 13: 155-200.<br>
> -Marsden, P.V., Friedkin, N.E. 1993. Network studies of social influence. S=
> ociological Methods&amp; Research 22, 127-151.<br>
> -Leenders, R.Th.A.J. 2002. Modelling social influence through network autoc=
> orrelation: Constructing the weight matrix. Social Networks 24: 21-47.<br>
> <br>
> Hope this helps,<br>
> Filip</span></font></p>
> <div>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;"><br>
> On 02-05-11,<b><span style=3D"font-weight: bold;">Jorge M Rocha</span></b=
>> &lt;<a href=3D"mailto:[log in to unmask]" target=3D"_blank">jrochagzz@GMA=
> IL.COM</a>&gt; wrote:</span></font></p>
> <div>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">***** To join INSNA, visit
> <a href=3D"http://www.insna.org" target=3D"_blank">http://www.insna.org</a>=
>   *****</span>
> </font></p>
> <table class=3D"MsoNormalTable" style=3D"" border=3D"0" cellpadding=3D"0">
> <tbody>
> <tr style=3D"">
> <td style=3D"padding: 0.75pt;">
> <div>
> <p class=3D"MsoNormal" style=3D""><font size=3D"3" face=3D"Times New Roman"=
>> <span style=3D"font-size: 12pt;">Dear all,</span></font></p>
> <p class=3D"MsoNormal" style=3D""><font size=3D"3" face=3D"Times New Roman"=
>> <span style=3D"font-size: 12pt;">I have a conceptual question that is prob=
> ably somewhat na=EFve regarding the analytic logic behind ERGM=92s (I=92ve =
> been using statnet and its suite of tools).&nbsp; If
>   someone in the community could point me in the right direction literature =
> wise, I would appreciate it.&nbsp; Now let me try to state my question/doub=
> t as briefly and clearly as I can:</span></font></p>
> <p class=3D"MsoNormal" style=3D""><font size=3D"3" face=3D"Times New Roman"=
>> <span style=3D"font-size: 12pt;"> From what I understand, the modeling fram=
> ework of an ERGM treats the observed network as the
> <u>dependent variable</u>  and the specified structural configurations and c=
> ovariate information about nodes/edges as
> <u>independent variables</u>  whose presence/absence increases the log-likel=
> ihood for the model.&nbsp; In other words, what we want to achieve in these=
>   models is to gain a better understanding of the types of processes that mi=
> ght have gone into generating a network
>   similar to the observed one; hypotheses are about this or that configurati=
> on or node/edge attribute and their effect on the pattern of social interac=
> tions represented in the network. If so, how would I go about testing hypot=
> heses concerning the effects of
>   social interactions on nodal attributes (I=92m particularly thinking of di=
> fferent types of opinions individuals might tend to have based on their int=
> eractions)?&nbsp;&nbsp;
> </span></font></p>
> <p class=3D"MsoNormal" style=3D""><font size=3D"3" face=3D"Times New Roman"=
>> <span style=3D"font-size: 12pt;">What I=92ve done so far is kind of arguin=
> g backwards (or at least that is how it feels to me =96maybe because I=92m =
> coming from a traditional linear/logit regression
>   background, and probably part of my problem is thinking in terms of DV and=
>   IV).&nbsp; If the coefficients are significant for some node covariate of =
> my interest in the ERGM estimation, then I=92ve been interpreting this as e=
> vidence that it is not unreasonable to
>   argue that the pattern of social interactions influences the type of opini=
> on individuals have (I know that if I had longitudinal data this issue of r=
> everse causality would not be so much of an issue, but so far all I have is=
>   cross-sectional data).&nbsp; How wrong
>   am I on this (the backward-arguing)?</span></font></p>
> <p class=3D"MsoNormal" style=3D""><font size=3D"3" face=3D"Times New Roman"=
>> <span style=3D"font-size: 12pt;">Finally, part of my worry is that I=92ll =
> be sending the research to a consumer behavior/marketing journal where the =
> familiarity with social network stuff =96let alone
>   ERGM=92s=97is limited.&nbsp; Any suggestions on how to best explain these =
> issues to a non-specialized audience?</span></font></p>
> <p class=3D"MsoNormal" style=3D""><font size=3D"3" face=3D"Times New Roman"=
>> <span style=3D"font-size: 12pt;">In advance, thanks,</span></font></p>
> <p class=3D"MsoNormal" style=3D""><font size=3D"3" face=3D"Times New Roman"=
>> <span style=3D"font-size: 12pt;">&nbsp;</span></font></p>
> <p class=3D"MsoNormal" style=3D""><font size=3D"3" face=3D"Times New Roman"=
>> <span style=3D"font-size: 12pt;">Jorge M Rocha</span></font></p>
> <p class=3D"MsoNormal" style=3D""><a name=3D"OLE_LINK2"></a><a name=3D"OLE_=
> LINK1"></a><font size=3D"3" face=3D"Times New Roman"><span style=3D"font-si=
> ze: 12pt;">EGADE</span></font>  Business School</p>
> <p class=3D"MsoNormal" style=3D""><font size=3D"3" face=3D"Times New Roman"=
>> <span style=3D"font-size: 12pt;">&nbsp;</span></font></p>
> <p class=3D"MsoNormal" style=3D""><font size=3D"3" face=3D"Times New Roman"=
>> <span style=3D"font-size: 12pt;">&nbsp;</span></font></p>
> </div>
> </td>
> </tr>
> </tbody>
> </table>
> <p><font size=3D"3" face=3D"Times New Roman"><span style=3D"font-size: 12pt=
> ;">_____________________________________________________________________ SO=
> CNET is a service of INSNA, the professional association for social network=
>   researchers (<a href=3D"http://www.insna.org" target=3D"_blank">http://www=
> .insna.org</a>).
>   To unsubscribe, send an email message to<a href=3D"mailto:[log in to unmask]
> ufl.edu" target=3D"_blank">
> [log in to unmask]</a>  containing the line UNSUBSCRIBE SOCNET in the bo=
> dy of the message.</span></font></p>
> </div>
> </div>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;"><br>
> -- </span></font></p>
> <div>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">&nbsp;</span></font></p>
> </div>
> <div>
> <p class=3D"MsoNormal"><font size=3D"2" face=3D"Arial"><span style=3D"font-=
> size: 10pt; font-family: Arial;">Dr. Filip Agneessens</span></font></p>
> </div>
> <div>
> <p class=3D"MsoNormal"><font size=3D"2" face=3D"Arial"><span style=3D"font-=
> size: 10pt; font-family: Arial;">Department of Sociology/ICS</span></font><=
> /p>
> </div>
> <div>
> <p class=3D"MsoNormal"><font size=3D"2" face=3D"Arial"><span style=3D"font-=
> size: 10pt; font-family: Arial;">University</span></font><font size=3D"2" f=
> ace=3D"Arial"><span style=3D"font-size: 10pt; font-family: Arial;">  of Gron=
> ingen</span></font></p>
> </div>
> <div>
> <p class=3D"MsoNormal"><font size=3D"2" face=3D"Arial"><span style=3D"font-=
> size: 10pt; font-family: Arial;">Grote Rozenstraat 31</span></font></p>
> </div>
> <div>
> <p class=3D"MsoNormal"><font size=3D"2" face=3D"Arial"><span style=3D"font-=
> size: 10pt; font-family: Arial;">9712 TG Groningen</span></font></p>
> </div>
> <div>
> <p class=3D"MsoNormal"><font size=3D"2" face=3D"Arial"><span style=3D"font-=
> size: 10pt; font-family: Arial;">The Netherlands</span></font></p>
> </div>
> <div>
> <p class=3D"MsoNormal"><font size=3D"2" face=3D"Arial"><span style=3D"font-=
> size: 10pt; font-family: Arial;">email:
> </span></font><a href=3D"mailto:[log in to unmask]" target=3D"_blank"><fon=
> t size=3D"2" face=3D"Arial"><span style=3D"font-size: 10pt; font-family: Ar=
> ial;">[log in to unmask]</span></font></a></p>
> </div>
> <div>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;"><a href=3D"http://filipagneessens.wordpress.com/" ta=
> rget=3D"_blank"><font size=3D"2" face=3D"Arial"><span style=3D"font-size: 1=
> 0pt; font-family: Arial;">http://filipagneessens.wordpress.com/</span></fon=
> t></a></span></font></p>
> </div>
> <p class=3D"MsoNormal"><font size=3D"3" face=3D"Times New Roman"><span styl=
> e=3D"font-size: 12pt;">____________________________________________________=
> _________________ SOCNET is a service of INSNA, the professional associatio=
> n for social network researchers (<a href=3D"http://www.insna.org" target=
> =3D"_blank">http://www.insna.org</a>).
>   To unsubscribe, send an email message to<a href=3D"mailto:[log in to unmask]
> ufl.edu" target=3D"_blank">
> [log in to unmask]</a>  containing the line UNSUBSCRIBE SOCNET in the bo=
> dy of the message.
> </span></font></p>
> </div>
> _____________________________________________________________________ SOCNE=
> T is a service of INSNA, the professional association for social network re=
> searchers (http://www.insna.org). 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.insna.org). To unsubscribe,
>   send an email message to [log in to unmask] containing the line UNSUBS=
> CRIBE SOCNET in the body of the message.</div>
> </div>
> </div>
> </body>
> </html>
> _____________________________________________________________________
> SOCNET is a service of INSNA, the professional association for social
> network researchers (http://www.insna.org). To unsubscribe, send
> an email message to [log in to unmask] containing the line
> UNSUBSCRIBE SOCNET in the body of the message.
>
> --_000_E3B21BD4B98F0E49905EF5CB1CE78695C92FMBXP09dsmanacuk_--

-- 
==============================================
Tom A.B. Snijders
Professor of Statistics in the Social Sciences
University of Oxford

Professor of Statistics and Methodology
Department of Sociology
University of Groningen
http://www.stats.ox.ac.uk/~snijders/
==============================================

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