2 Research Context
5.9 Limitations of Methodological Approach
5.9.2 Specific Issues Related to Ego Network Analysis and In-depth Interviews
The sample for the in depth interviews was not statistically representative of the population as is generally the case with qualitative methods of data collection. The purpose of utilizing SNA in combination with in depth interviews in the present study was to discover the meaning and understanding of norm salience rather than to verify or predict outcomes. Some qualitative experts argue that qualitative findings can be generalized to other people, settings, times and treatments to the degree to which they are similar to the people, settings, times and treatments in the original study (Johnson and Christensen, 2007). Stake (1978) uses the term ‘naturalistic generalization’ to refer to this process of generalizing qualitative explanations on the basis of similarity. Whilst it is accepted that the network results and related qualitative interpretations uncovered in this study may not be generalized to the population in traditional sense of the word, it is possible that the study may have uncovered a phenomenon which is particularly related to commuter colleges (More on this in chapter 9, section 9.5). Therefore, the findings may be tentatively and naturalistically generalized to the extent of being especially relevant to commuter colleges.
Self Reported Network Data
One of the inherent features of ego network analysis relates to the process of determining network structure. Unlike whole networks, which are based on a near-complete enumeration of the population of interest and collect data from every member of the network, ego networks view social environments from the eyes of the focal individual (ego). Egocentric network data is therefore based purely upon the knowledge, reflection and recall of the ego (O'Malley et al., 2012). It captures the diversity in the social environment of individuals and it is because of this diversity that the boundaries of ego networks are dependent on the context being examined and
generally not known at the onset of an inquiry. Since it is not possible to collect network data from each and every member of an ego network, egos (participants) report on the presence or absence of ties between each pair of nodes in their networks. The possibility that participants might not have been aware of all possible ties in their networks or might not have recalled them fully is thus inevitable. Network analysts believe that examining ego networks is still relevant and useful considering that an ego’s perception of relationships may be more important than whether or not the perceived relationship is validated via reciprocation by the alter (O'Malley et al., 2012).
Missing Data
Social network studies are especially sensitive to missing data which often occurs due to non response. It can either take the form of a participant leaving out a relevant node altogether referred to as unit non-response or not completing specific items related to a node(s) referred to as item non-response. Researchers caution that this can have negative effects on the structural properties of networks (Burt, 1987; Ghani et al., 1998; Borgatti and Molina, 2003b; Kossinets, 2006). The problems associated with sampling and missing data in whole network studies often stem from the inability of researchers to interview or observe network members. For ego networks, alters and ties are often missing because respondents either did not recall them or were not asked about them in such a way so as to fully capture the network structure. Several steps were taken to minimize the occurrence of missing data in this study. First, free recall method was used to extract ego networks which is argued to be the most comprehensive though time consuming method of eliciting network members (Ferligoj and Hlebec, 1999; Carrasco et al., 2008). Second, multiple name generators were administered to ensure inclusion
of all relevant nodes. In addition, a third name generator was administered probing if there was anyone else that the participant might have missed in the first two questions and would want to mention. Third, item non response was addressed by reviewing the alter attribute chart, the concentric circles evaluation and the adjacency matrix prior to the termination of each interview and complete any missing information. This step ensured that there was no missing data with regards to alter attributes, tie strength scores and interrelationships of alters. Finally, the understanding of the functionality of ego networks and the formation, dissemination and reinforcement of drinking norms within them was based on a qualitative interpretation of the interviews which strengthened the analysis and further reduced the chances of omitting socially relevant and salient nodes.
Researcher Bias
As with any qualitative inquiry, researcher bias might have occurred and therefore it is acknowledged. Researcher bias is an important concern in qualitative research because this type of research tends to be exploratory, open ended and less structured than quantitative methods (Johnson and Christensen, 2007). It may occur in the form of selective observation, selective recording of information and also from allowing one’s personal opinions and perspectives to affect how qualitative data is interpreted and how the research is conducted. While it is not possible to completely eliminate researcher bias from any qualitative study (Johnson and Christensen, 2007), efforts were made to reduce its impact by improving the rigor of data collection in keeping with research on qualitative methods (Johnson and Christensen, 2007). This was achieved through the use of neutral probing, fewer assumptions and avoidance of leading questions and/or premature interpretation. In addition, the use of low
inference descriptors was maximized in deriving qualitative explanations and reporting them. This was achieved by phrasing descriptions very close to the participants’ accounts and researcher’s interview notes and using direct quotations where appropriate to expand on the findings.
6
Analysis and Results of Survey Data
This chapter presents the results of analysis performed on the survey data and serves the overarching aim of addressing objectives one, two and three of this research as outlined in chapter 5, section 5.2. It begins with developing a feel about the data and presenting the key descriptive findings which establish an understanding of alcohol consumption levels across DIT. Self reported personal consumption is then compared with perceived peer norms to assess if students misperceived these norms as has been found in prior studies described in chapter 3, section 3.6.2. The results of hierarchical multiple regressions follow. These results address two important aspects of this study. First, the effect of perceived peer norms of different referent groups on personal consumption is investigated. Second, the relative impact of perceived descriptive and injunctive norms on personal consumption is examined. The chapter concludes with a short discussion of key findings.