«NETWORKING IN EVERYDAY LIFE by Bernard J. Hogan A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy ...»
networking in different ways with those they know in person. So I should ﬁnd that:
Hypothesis 5 (On personal characteristics): Recent immigrants will show signiﬁcantly more distinctive networking strategies with their network by virtue of their need to sustain both local and long distance ties.
This sets up a relatively straightforward model using a series of distinct aspects of one’s social network and social context to predict increased person-to-person networking. But how can I condense the many forms of networking into a single variable amenable to a single model?
Here I introduce a novel concept—the particularity of media use. This concept refers to the idea that some individuals have very particular and person-speciﬁc media use behaviours, both in terms of frequency and the number of media used, while individuals have very general strategies. A general strategy implies that one networks the same way with all of their alters, while a particular strategy implies that one networks in unique and speciﬁc ways depending on the tie.
Barbara and Priscilla, two individuals from the interviews, offer an interesting descriptive counterpoint on the process of particularity in action. Moreover, both networks are almost topologically isomorphic (i.e., they look almost the same when drawn as a network, see Figure 6.2). Finally, these networks were chosen because they represent polar ends of a particularity score, which is described in the next section.
CHAPTER 6. WITHIN-NETWORK VARIATIONS AND NETWORKED INDIVIDUALISM 157An example of particularity Both Priscilla and Barbara are young women with children. Priscilla, aged 26, has two children. She is separated from her husband who is in the military and currently overseas. Barbara is somewhat older at 36 years of age. She lives with her partner and they have three children together. Both networks have 14 alters, a triad and the rest are isolates plus a few dyads. Both might be considered networked individualistic structurally. Their networks are sparse, far-ﬂung geographically and loosely connected.
Yet, these two cases are on the opposite extremes of media particularity. Priscilla had the highest particularity score and uses a seemingly different strategy with almost all of her alters. Moreover, her score is ampliﬁed by the fact that she does not use different strategies occasionally, but on a routine, almost daily basis. Barbara had the second lowest particularity score. She uses a single medium (primarily telephone) with almost every network member, plus in-person interaction with a few friends. A key difference here is that Priscilla uses a combination of media with any given person, whereas Barbara seems to prefer the ‘right’ way to contact people, and uses only that way.
Further inspection of the interviews reveals that Priscilla’s strategy is only partially her doing. While she is very particular, part of this particularity is not because she actively calls people via a variety of media, but because she makes herself accessible by a variety of media. In the interviews we asked if individuals contact alters, or the alters contact the individual. In Priscilla’s case, many individuals would contact her, rather than be symmetric or wait for her to contact them. She notes that others simply show up at the house (to visit the kids or help), or the other person initiates email or an instant messenger conversation. It’s not that way for all members, but for most. Here one can see how Priscilla is taking advantage of many affordances of media (asynchronicity as well as long distances) by regulating access rather than being overly active with her alters. By contrast, Barbara overwhelmingly claimed that
CHAPTER 6. WITHIN-NETWORK VARIATIONS AND NETWORKED INDIVIDUALISM 158
Figure 6.2: Radial pie chart networks showing variations in contact frequency.
Grayed out nodes indicate missing data.
CHAPTER 6. WITHIN-NETWORK VARIATIONS AND NETWORKED INDIVIDUALISM 159all of her alters contact her about as frequently as she contacts them. For her, it is not about regulating access so much as maintaining stability. She prefers the telephone, and prefers to use it with people who keep up (i.e., calling and receiving calls equally).
As one can see from these two networks, Priscilla is far more active than Barbara, but that activity is also partly about habit. By routinely checking email, leaving her instant message account active and keeping her telephone with her, she is regulating access so that other people can use what strategy suits them best for contacting Priscilla. By contrast, Barbara says she tried instant messaging once and did not like it. She imbues media with affect rather than affordances. For her it is ‘less than’ in person or voice contact, rather than being convenient or useful.
6.4.2 The particularity score Moving from two networks to the entire data set, I need to quantify particularity in some meaningful way. There are numerous means for testing the differences between individuals in a given set, but few of these techniques allow for large scale comparisons across sets. If I was examining one or perhaps a few networks, comparing the networks using multidimensional scaling would be a sensible option. For each network, I could plot the differences in media use between networks. It would be a visual diagram that lays out the members of a network according to their differences in media use. If all members of the network use media similarly with ego then they would all be clumped together. If ego has two clear but distinct strategies (say one with workmates and one with family members), then there would be two clumps.
However, 86 multidimensional scaling plots are no way to interpret these personal networks. Fortunately, I can draw upon the logic of multidimensional scaling to get an overall assessment and thus a ranking of individuals by media use.
Fundamentally, multidimensional scaling calculates pairwise comparisons of variables for every case. So if there are 4 cases, it calculates 6 (or (n*(n-1))/2 ) comparisons.
CHAPTER 6. WITHIN-NETWORK VARIATIONS AND NETWORKED INDIVIDUALISM 160The comparisons generally use a dissimilarity measure such as Euclidean distance (which is also used to calculate distance in the common network analysis technique of blockmodeling). In the standard form, the formula for this distance is very straightforward as the root of the sum of squares of the differences between variables. For two points P and Q in N -dimensional vector space (i.e., two cases each with values
on n variables), the Euclidean distance is calculated as:
Thus, if two nodes in a network interact with ego in similar ways then they will have a low distance. By contrast, if nodes show different trends, either because they use the same media in different frequencies or use different media, then their distance will be high in this space. Traditionally, one would use Euclidean distance as a means for partitioning or interpreting a single data set. However, in this case, I need to not only ﬁnd the distances for a single network, but somehow compare this to all the distances for another network. One option would be to present and visually inspect 84 multidimensional scaling plots. But a more parsimonious option would be to somehow distill these distances into a measure comparable across networks. I term this measure the particularity score.
This particularity score is simply the average of all the pairwise distances in the network. To note, this is not the average of the number of cases, but the number of pairwise comparisons. For n cases, there are (n(n − 1))/2 comparisons.
If the particularity score is high, then one can infer that there is a great deal of distance between all the cases in the network.2 But if the score is low one can infer that there is a great deal of homogeneity in the network. The score is obviously dependent on the sort of variables that are put into the formula. If one uses a simple ordinal scale To be fair, recall media use measurements are not taken for the entire network, but only those network members for whom we administered a minisurvey. Further details about these cases are found in Chapter 4 as well as Hogan et al. (2007).
CHAPTER 6. WITHIN-NETWORK VARIATIONS AND NETWORKED INDIVIDUALISM 161where 1 = yearly, 2 = monthly, 3 = weekly and 4 = daily, it will produce a somewhat different result than a score that is converted to days where 1 = yearly, 12 = monthly, 50 = weekly and 300 = daily. I opted for the ordinal score since represents a compromise. Simply looking at differences in use/non-use, would be too broad to allow us to distinguish active ties from one-time ties.
The particularity score calculated through this process is a normally-distributed set of values for all valid respondents.3 The value is not immediately intuitable as a point estimate. But simply stated, it represents the average Euclidean distance between alters on media use with ego. The distribution of this value is presented in Figure 6.3 Percent of respondents Figure 6.3: Distribution of particularity score for differences of media use in personal networks I believe that the particularity of media use is a good estimate of networked individualism, as it suggests that individuals have particular media use strategies by alter rather than general strategies for the entire network. Whereas in the last chapTo be more precise, I can reject the null hypothesis when testing for normality using the ShapiroWilk test (p = 0.614). To note, in this test one is looking for a non-signiﬁcant p-value.
CHAPTER 6. WITHIN-NETWORK VARIATIONS AND NETWORKED INDIVIDUALISM 162ter I clustered the sample on overall media use, in this chapter I consider differences within networks. One key difference, thus, between this chapter and the last is that this chapter uses a more restricted version of the data set, namely those individuals who completed both a survey and an interview (N = 86) rather than the complete sample (N = 350).
In this work, I am interested in the overall dissimilarity within each network, and how such dissimilarity compares to other networks. Basically, where there is a lot of dissimilarity, it is because there is a lot of variation in media use. Where there is little dissimilarity there is little variation in media use. By deﬁnition, the particularity measure captures this dissimilarity.
The dissimilarity value is not at all correlated with the size of the network, although it is correlated with the frequency of contact. Most notably with face-to-face and email contact, suggesting the frequency of these two values demonstrates the greatest differences among the networks.
6.4.3 Predicting particularity A point estimate of the particularity of media use in these networks accomplishes a number of things. The ﬁrst is simply a shortcut to assessing the difference in media use among networks. For example, I used the dissimilarity measure when assessing which networks to select for the comparison shown above. Second, it works as a possible covariate for subsequent analysis of networked individualism, i.e., do people who score high on particularity also score high on certain network measures? Third, this value can work as a response variable—namely, what other factors can explain the variation in particularity?
The challenge of this score is whether or not the reader is persuaded that it represents something real, or is merely a statistical ﬁction. I believe that this measure is reasonable for several reasons. The ﬁrst is that the formula is normalized across netCHAPTER 6. WITHIN-NETWORK VARIATIONS AND NETWORKED INDIVIDUALISM 163 works of varying size, and that the pairwise calculations are intelligible (it is really just the addition of the differences in media use between two alters, for all valid alter pairs). The second reason is that it has a certain face validity. When I examine the networks with the highest and lowest particularity score, patterns of media use become apparent. Thus, it is an illuminating way to order the networks. The third reason is that conceptually this measure scales well with the possible addition of new media or new cases. It is the sort of measure that can be reapplied to subsequent studies of media use to examine intra-sample differences, so long as one collects network data.
Since this is a new measure, there is little literature that can act as guidance about what other variables can help explain particularity. However, there are logical arguments justifying the use of some measures rather than others.
1. Social location: Numerous social location measures can be posited to explain particularity. Perhaps men are more general and consistent while women more sensitive to differences among alters. Or perhaps individuals with family members living abroad must be more particular if they want to access the remaining family members.
2. Media use: Obviously, since using the Internet is a pre-condition for accessing email and instant messaging (which are themselves a means to increased particularity), Internet use should ﬁgure into a model of particularity. Apart from that is the frequency of communication as well as the number of individuals one plans with by any given media. I will use the latter measures with care since frequency of media use, in some form, is how the particularity measure was derived in the ﬁrst place.
3. Network structure: Ultimately, the goal of this measure is to assess whether individuals who have particularistic networking styles also exhibit networked individualistic network structures. Thus, centralization, density, the number of alters, the size of the largest component and the number of isolates should all be candidates for inclusion in a model.