«NETWORKING IN EVERYDAY LIFE by Bernard J. Hogan A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy ...»
5.3.2 Clustering II: Interpreting the cluster Results of the k-means analysis of the six variables for planning indicated that a sixcluster solution was optimal. For this analysis, the score for the Calinski-Harabasz gradually decreased as expected, but there was a steep drop in the size of the largest cluster between k = 5 and k = 6. Since there is only a slight decrease in the score, but a very large redistribution in cases, it appears to be the most parsimonious solution.
Figure 5.1 plots these values across a range of k partition solutions.
The six different partitions are referred to by the amount of planning done through the most dominant media. They are as follows: “Heavy All Media”, “Heavy Cell and The complete formula is (SSB/(k − 1))/(SSW/(n − k)) where n is the number of data points and k is the number of clusters. SSB stands for the sum of squares between the partitions and SSW stands for the sum of squares within the partitions.
CHAPTER 5. NETWORK PROFILES: COUPLING MEDIA USE AND SOCIAL ACTIVITY 103Calinksi-Harabasz Score
Figure 5.1: Calinski-Harabasz scores and largest partition size by values of k Face-to-Face” (hereafter F2f), “Heavy Telephone”, “Heavy F2f”, “Moderate Telephone and F2f”, “Light planning”.
I have ordered these groups by their total mean amount of planning (where “Heavy all” obviously does the most planning overall, and “Light planning” does the least. Neither group was particularly heavy in planning by email, cell phone (texting) or instant messaging, although the “Heavy all” group used email as frequently as cell phones. Also, it is worth noting that there was a lot of variation in the secondary media used in the “Heavy F2f”, so that some of these individuals supplemented a lot of in person planning with email use, while others supplemented it with a lot of instant messaging use (and others, more conventionally used the telephone).4 In a sense, these media use partitions are media use “styles”. These partitions This can be contrasted with the idea that people in this group used all media lightly but evenly— that was not the case. This was discovered through analyses of individual values as well as an assessment of the standard deviations of the mean planning values.
CHAPTER 5. NETWORK PROFILES: COUPLING MEDIA USE AND SOCIAL ACTIVITY 104
represent a coherent set of media use habits. Each partition indicates a dominant medium (or media) that is consistently used in a routine fashion alongside secondary media that are used with greater variance. By virtue of having a medium “in one’s toolkit” an individual has a particular level of access to alters, and is accessible by alter in a particular way.
These proﬁles do not represent absolute styles as there are countless other factors that are embedded in one’s habitual media use. Nevertheless, these partitions differ substantially from each other and display a clear internal logic. For all partitions, planning in person and planning via the telephone ﬁgure prominently, although they decrease in frequency from one partition to the next.
Recall that these measurements were taken from an index of media use per month as there were two separate measurements per media, one for very close alters and one for somewhat close alters. Individuals certainly vary by the proportion of very close alters as well as the proportion of somewhat close alters in their networks. Granted, it is possible to look at even more complicated cluster solutions by splitting up (and
CHAPTER 5. NETWORK PROFILES: COUPLING MEDIA USE AND SOCIAL ACTIVITY 105maybe weighting) the media use by somewhat close and very close. However, such additional analyses are considered outside the scope of this analysis since I am investigating an overall propensity to plan, not how clearly people demarcate who is the stronger ties (i.e. the very close alters) and the weaker ties (i.e. the somewhat close alters).
One thing one can notice in this chart is that as the clusters increase in activity, they decrease in size. That is, the largest partition is the substantial chunk of individuals who make plans with their personal network on a weekly basis or less. By contrast, the smallest partition is the one that has the most active planning habit. Selecting such a small distinct group (N = 9) is partially a quirk of the k-means algorithm, but oneway ANOVA tests conﬁrm that the six groups differ from each other on frequency of use of all ﬁve media plus face-to-face (p ≤ 0.001). Furthermore, post-hoc Bonferroni tests examining pairwise differences reveal that almost all of the partitions differ from each other on every medium, rather than having a single partition be responsible for the differences. Only in the case of instant messaging and texting by cell phone were several of the partitions not signiﬁcantly different from each other.
As a consequence of this analysis I can say that those nine heavy media users really are distinct from the rest of the sample in terms of their active use of media. Also, the analysis indicates that there really is not that much distinguishing the 147 individuals in the largest and least active partition. Recalling Figure 5.1, I would have to double the number of partitions from k = 6 to k = 12 in order to dramatically affect the size of that largest partition, and even then, the largest two partitions are still substantively similar. What is more important is that the small group of nine “media omnivores” appear to do far more than their fair share of planning. I will return to this in the section below, as I suggest these individuals are probably the social hubs that link different groups together, and are generally very adept at “networking”, in the common senses I referred to at the beginning of Chapter 2. For example, they do not differ signiﬁCHAPTER 5. NETWORK PROFILES: COUPLING MEDIA USE AND SOCIAL ACTIVITY 106 cantly in any of the demographic or social location variables—they are not a wealthy networking elite, but rather are distributed throughout various social locations and life courses. The one thing uniting them is their compulsion to be active with many network members in whatever way possible.
Another noteworthy ﬁnding is that there is no group that solely makes plans in person. The ‘Heavy F2f’ partition clearly does most planning in person, (mean 31.2 times per month), yet these individuals also use the telephone at least twice a week on average. Also, not a single case in this partition uses in person contact exclusively. It seems that if one is to be an active planner, it must involve at least some mediated contact.
Furthermore, this “Heavy F2f” partition is second only to the media omnivores in their use of instant messaging. Only 9 of the 23 members in this group use instant messaging, but these 9 individuals use instant messaging at least three times a week on average.
5.3.3 Clustering III: Interpreting the social locations of the clusters One question about networking styles is the extent to which they are reactionary responses to particular social situations, or more internally governed states of communication. For example, do individuals with more income plan by more media? Do women plan more then men, or older individuals plan using more traditional means than younger individuals? Table 5.2 shows the mean differences in numerous social location variables, as well as network size, and total propensity to plan. As can be seen from this table, the clusters do vary signiﬁcantly on a number of variables. However, the relationship to social location is weak, as evinced by the lack of pairwise significant differences and the low signiﬁcance of the ANOVA scores. If the relationship was strong, not only would the ANOVA model be signiﬁcant, but pairwise comparisons within the ANOVA would also be signiﬁcant.5 Nevertheless, media use styles Pairwise comparisons were made using Bonferroni’s post-hoc test.
CHAPTER 5. NETWORK PROFILES: COUPLING MEDIA USE AND SOCIAL ACTIVITY 107do seem to vary by age, whether individuals have children at home and whether they are coupled. In general, older individuals, those with children and those who are coupled have different styles than those who are younger, single, and childless.
Younger individuals clearly do more planning than older individuals, all else equal.
However, the results are not as simple when considering single versus coupled and children versus childless. Relatively young single individuals are more likely to be mobile and plan using cell phone than those with children and couples. However, the media omnivores actually have the most children at home, and yet are less likely to be coupled than most of the clusters. This paints an interesting picture. Younger individuals are more prone to use a diversity of media. However, when these individuals have children—especially single mothers with children, they are prone to using these devices intensively to coordinate action with their network. To note, three of the media omnivores are single mothers, while the remainder are any combination of gender, having children and marital status.
Also, there is a clear signiﬁcant negative correlation between age and total planning frequency (r = −0.27, p 0.001). Interestingly, however, there is also a curvilinear effect of age. Those who are retired plan substantially more than their work-age brethren. This pattern is shown in Figure 5.3. The ﬁgure is a scatterplot of the number of times individuals plan per month by all media over age. Superimposed over this scatterplot is a quadratic (i.e., curved) trend line. Surrounding this trend line is the 95 percent conﬁdence interval for the trend. This reﬂects a general life course tendency for individuals to be most active in planning when they are younger, decrease as they settle into middle age and child-rearing and then increase somewhat as children move
people adopt more technology more rapidly and plan more frequently. The second trend is that planning is constrained by the amount of time people have to do it. Middle age, being the most time-scare and harried period of life (Southerton, 2003; Robinson and Godbey, 1999), is also the time when there is the less social planning outside the home.7 Individuals who are older still are less likely to be constrained by children, even as they become increasingly physically constrained.
Instances of planning per month (all media)
Beyond age and time, there is one other very signiﬁcant factor differentiating media use styles herein. The networks of the small group of media omnivores are particularly large. Overall, what distinguishes the media omnivores is not whether they have children or are single. Indeed, there are more single female parents in this group than one might expect by chance but since the cluster only contains nine individuals it is hard to conﬁrm this result. The real difference is in the size of the network. The omnivores plan and are very busy, but this planning is clearly directed at engaging a It is worth recalling here that these ﬁgures are for planning with alters, in this case meaning individuals who the respondent voluntarily associates with and who do not live with the respondent.
Middle-age, by contrast is a harried time partially because of the demands within the family, something which is not explicitly covered in this analysis.
CHAPTER 5. NETWORK PROFILES: COUPLING MEDIA USE AND SOCIAL ACTIVITY 110substantially large swath of individuals, and by far and away a larger swath than one would ﬁnd by chance.
By contrast, the network sizes of the other ﬁve groups do not differ signiﬁcantly.
One might say there is a substantive difference between the “Heavy F2f” partition and the others, since they have more very close alters, but their total number of alters is about the same as the other partitions (omnivores notwithstanding). Thus, one can see, that in East York, there is a general network size of about 35 members (using the summation method), with about 75 percent more somewhat close alters than very close alters. One’s media use style does not seem to have a big impact on the number of alters, except in the case of the media omnivores who appear to use any medium at any time with many alters.
From the qualitative interviews, what stands out the most about the heavy media users is not their consistent social location, but the fact that participants in the heavy media group tend to have a blended home-work situation. For example, one is a student and single mother (#243 “Priscilla”), another is a self-employed concert promoter and sales director (#601 “Clay”), the third is a nutrition consultant (#431 “Wendy”) and a fourth works two jobs, one as a PR person out of a home ofﬁce (#672 “Hedda”). All four talk about how important email and cell phones are to set up meetings. The nutrition consultant was particularly sensitive to the complexities of scheduling and the
useful affordance of multiple email recipients: