***** To join INSNA, visit http://www.insna.org *****
You've described a few ways to operationalize "ego tie strength"
(procedures you might use to measure it) but there didn't seem to be
many words about your conceptualization -- what do you mean by "ego tie
strength"? The problem is that if you don't know what you are trying to
measure, you will never know if you are measuring it.
Unless, that is, you say the concept is defined by the procedure used to
measure it. A conceptual definition must do more than tell how the
concept will be measured. For example, to say that “intelligence is
defined as that which the IQ test measures” neither defines intelligence
nor tells you anything about it. It only ties the word "intelligence" to
something called the "IQ test".
The most important part of a conceptual definition is the specification
of essential qualities. Conceptual definitions should denote all of the
essential qualities of the constructs and exclude nonessential ones.
They should describe the construct clearly enough so that other
researchers would classify phenomena (in terms of whether or not they
are occurrences of the construct) the same way as the researcher who
developed the conceptual definition.What are the essential qualities of
the concept? In your case, what thing or combination of things must
happen when "ego tie strength" is high? When that thing or combination
of things doesn't happen, "ego tie strength" should be low.
The definition “violent movies are movies in which there are scenes of
kicking, stabbing, clubbing, choking, hitting, slapping, shooting, and
smashing” is not a good conceptual definition because it lists many
activities that are not essential and it fails to describe the essential
qualities. What are the conceptual qualities that must be present before
you will call something “violent”? You may want to say something like
“violent movies show scenes in which one person injures, maims, or
causes pain to another.” This much shorter list of more abstract
qualities would probably include all of the specific activities in the
list above as well as many violent acts not on that list.
The specification of essential qualities is the most important part of a
conceptual definition because it gives some very good clues about how
the concept could be measured in a most straightforward way: look for
the presence or absence of the essential qualities. For example, if you
accepted the suggested definition of violent movies as those which “show
scenes in which one person injures, maims, or causes pain to another” in
the previous paragraph, you could determine whether or not a movie is
violent, by looking to see whether it contained scenes in which one
person injures, maims, or causes pain to another. ... But first you will
have to decide what counts as injury or pain. (The proceeding was taken
from Zen of Empirical Research; see www.sfu.ca/~richards)
In a similar way, the following may specify procedures you could use to
get a number for each of a set of egos (but none of them tells you
anything about what "ego tie strength" is): the mean of the closeness
ratings that ego gives to his or her social ties, the mean of the
products of the closeness and frequency ratings that ego gives to his or
her social ties, the mean rank of ego's top 10 ties, the mean rank of
ego's top 11 ties, .... , the result of some other mathematical
procedure that somehow produces a single number that summarizes some
rating that ego gives to some number of connections with a set of alters
chosen by some yet to be specified procedure, and last but not least:
ego's hat size multiplied by the number of dogs ego has owned in the
past eleven years. No doubt the last example on the list seems to be a
strange way to measure "ego tie strength," but we don't know what "ego
tie strength" is, so we can't really rule it out, can we?
I think this is not an unusual situation with network research. A lot of
people have spent a lot of time inventing a large set of metrics,
scales, indices, etc., which are supposed to measure some network
characteristic or another. We've got a lot of computer programs and
algorithms to use in our efforts to analyze networks. It's easy to pick
the tool you like and start analyzing. You'll get numbers. Because
social networks are very complex things (they're a lot more than an
adjacency matrix or the world's most beautiful multidimensional
sociogram -- social networks have people with histories, memories,
plans, dreams, desires, urges, diseases, jobs, families, fears, beliefs,
friends, enemies, ......) you will probably be able to come up with an
explanation for your numbers that makes sense. Worse, you will probably
be able to come up with an explanation that will make absolutely good
sense, regardless of what numbers you get.
So please spend at least a few hours trying to figure out exactly what
you mean by "ego tie strength." While you're doing that, do everything
you can to avoid numbers and mathematical procedures that turn one set
of numbers into another number or set of numbers. What you're interested
in is "what is ego tie strength?" -- not "how do I measure ego tie
strength?" When you know what it is, write it down. Write it clearly
enough so that someone who reads it will understand exactly what you
mean by ego tie strength. Then you will have less trouble coming up with
a way to measure it as long as you keep your conceptual understanding
clear in your mind.
all the best,
> Date: Thu, 7 Jun 2007 11:35:34 +1000
> From: Kenneth Chung <[log in to unmask]>
> Subject: Measure of Ego Tie Strength
> Dear colleagues,
> I am seeking ideas and suggestions on how one accounts for tie strength of
> an ego node, when you are studying over 100 ego nodes individually. In my
> study, relational data is collected from over 100 individuals, where each
> individual may elicit up to maximum 15 ties. The research model tests
> whether tie strength is associated with individual outcome, eg. attitude to
> Tie strength in my study is measured by:
> - closeness (4 point scale from very close to distant) and
> - frequency of contact (5 point scale ranging from daily to less often),
> although data on other variables such as 'time known' and 'relationship
> type' is also available.
> When it comes to calculation of tie strength for an ego node, how does one
> account for it? To the best of my knowledge and from what I've read from
> literature, one may
> 1. use the average strength of ties for an ego (ie. sum the values of each
> tie from ego to alter and divide by count of ties). In this case, the values
> of each tie may be:
> (i) the average of closeness and frequency values, or
> (ii) the product of closeness and frequency values
> 2. using 1, but take only the average of the top 5 or top 7 ties to the ego.
> This allows for comparison using a common baseline.
> Other approaches have been to consider tie strength of a node in terms of
> network proportions (see Reagans, R., & McEvily, B. (2003). Network
> Structure and Knowledge Transfer: The Effects of Cohesion and Range.
> Administrative Science Quarterly, 48, 240-267.), however, I understand this
> only works for sociocentric networks and not for ego networks as in my
> I'd like to confirm whether my limited understanding is correct and welcome
> comments and suggestions from you all.
> Thank you,
> Kon Shing, Kenneth Chung
> PhD Candidate
> School of Information Technologies
> University of Sydney
> NSW 2006, Australia
> P: +61 2 9351 5639
> F: +61 2 9351 3838
> W: http://www.it.usyd.edu.au/~ken
> Date: Thu, 7 Jun 2007 08:44:56 -0700
> From: Tom Valente <[log in to unmask]>
> Subject: Re: Measure of Ego Tie Strength
> You should treat tie strenght as a dyadic value.
> Kenneth Chung wrote:
>> Dear Tom,
>> Thank you very much for your input.
>> If I understand you correctly, assuming that an ego elicits 10 alters, I'd
>> rank the first alter a tie-strength-value of 10, with the last a value of 1.
>> To then arrive at a tie strength (well, a mean value) for the ego, I'd then
>> sum up the values of all ties in this case (10...1=55), and then divide it
>> by ego's network size (ie. 10) yielding a mean ego tie strength of 5.5. Is
>> this correct? If so, it seems to me that tie strength for the ego in such
>> case is simply a function of how many alters are elicited or network size.
>> ie. 10 alters would always yield 5.5, 11 alters yields 6 (66/11), and so on.
>> Essentially, to test my research model, it seems necessary to arrive at a
>> tie strength for each ego by consoloditating the tie strength of all other
>> ties and averaging them. Burt's approach is at the dyadic level and relative
>> to network proportions of the ego. Using this approach in my study means an
>> ego will always have a tie strength of 1.
>> I wonder if my notion of averaging tie strength and drilling it down to a
>> single value for the ego is an incorrect conceptualisation. Any pointers to
>> further papers/ideas would be of great assistance!
>> Kind regards,
>> -----Original Message-----
>> From: [log in to unmask] [mailto:[log in to unmask]]
>> Sent: Thursday, 7 June 2007 9:38 PM
>> To: Kenneth Chung
>> Cc: [log in to unmask]
>> Subject: Re: Measure of Ego Tie Strength
>> I've used the rank order of the nominations as a proxy for tie strength
>> since spending time with someone may not necessarily equate with tie
>> strength. So I treat the first person named as stronger, than the second,
>> and second stronger than third, etc. We find this to be correlated with risk
>> behavior, people tend to engage in riskier behavior with their closer ties.
>> This can be done in both ego and socio-metric studies.
>> - Tom
>> Kenneth Chung wrote:
>>> Dear colleagues,
>>> I am seeking ideas and suggestions on how one accounts for tie strength
>>> of an ego node, when you are studying over 100 ego nodes individually.
>>> In my study, relational data is collected from over 100 individuals,
>>> where each individual may elicit up to maximum 15 ties. The research
>>> model tests whether tie strength is associated with individual outcome,
>>> eg. attitude to performance.
>>> Tie strength in my study is measured by:
>>> - closeness (4 point scale from very close to distant) and
>>> - frequency of contact (5 point scale ranging from daily to less
>>> often), although data on other variables such as 'time known' and
>>> 'relationship type' is also available.
>>> When it comes to calculation of tie strength for an ego node, how does
>>> one account for it? To the best of my knowledge and from what I've read
>> >from literature, one may
>>> 1. use the average strength of ties for an ego (ie. sum the values of
>>> each tie from ego to alter and divide by count of ties). In this case,
>>> the values of each tie may be:
>>> (i) the average of closeness and frequency values, or
>>> (ii) the product of closeness and frequency values 2. using 1, but
>>> take only the average of the top 5 or top 7 ties to the ego.
>>> This allows for comparison using a common baseline.
>>> Other approaches have been to consider tie strength of a node in terms
>>> of network proportions (see Reagans, R., & McEvily, B. (2003). Network
>>> Structure and Knowledge Transfer: The Effects of Cohesion and Range.
>>> Administrative Science Quarterly, 48, 240-267.), however, I understand
>>> this only works for sociocentric networks and not for ego networks as
>>> in my study.
>>> I'd like to confirm whether my limited understanding is correct and
>>> welcome comments and suggestions from you all.
>>> Thank you,
>>> Kon Shing, Kenneth Chung
>>> PhD Candidate
>>> School of Information Technologies
>>> University of Sydney
>>> NSW 2006, Australia
>>> P: +61 2 9351 5639
>>> F: +61 2 9351 3838
>>> W: http://www.it.usyd.edu.au/~ken
>> Evaluating Health Promotion Programs (Oxford U. Press):
>> My personal webpage:
>> The Empirical Networks Project
>> Thomas W. Valente, PhD
>> Director, Master of Public Health Program http://www.usc.edu/medicine/mph/
>> Department of Preventive Medicine School of Medicine University of Southern
>> California 1000 S. Fremont Ave.
>> Building A Room 5133
>> Alhambra CA 91803
>> phone: (626) 457-6678
>> fax: (626) 457-6699
>> email: [log in to unmask]
Bill Richards, Ph.D.
School of Communication, Simon Fraser University
8888 University Drive, Burnaby, B.C. Canada V5A 1S6
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.