Drawing or tabulating ego-centered networks? A mixed-methods comparison of questionnaire vs. visualization-based data collection

Abstract

There is an increasing body of research which examines the reliability and validity of different modes of network measurement with most studies focusing on the classical paper or digital network questionnaires. Visual methods of data collection, in which interviewees draw their ego-centered network in a pictorial representation, still live a kind of shadowy existence in the field of network research, with little known about data quality of these approaches. In this paper, the classic approach of questionnaire data collection will be compared to the digital-visual method (VennMaker) in a parallel study design. Similar name generator and interpreter questions were used to collect network data from 264 young adults. Two participants from both studies participated in follow-up qualitative interviews to understand the advantages and disadvantages of data collection modes. Network measures of size, composition and density with additional information on the subjective importance of relations within the network were collected. Results indicate relevant differences between both network methods. These emerge due to the differential effect of the name generator in the visual approach, the specific display of the study material, and the particular limitations of space and effort each method entails. In-depth interviews shed light on subjective strategies (such as gestalt and distancing-and-selecting strategies) in order to cope with the distinct methodical tasks.

Keywords:

• Social network analysis
• visual data collection
• network questionnaire
• comparative approach
• mixed-methods

Acknowledgements

We want to thank the three anonymous reviewers and Ruth Lowry (University of Chichester) for their insightful comments that substantially helped to improve this paper. We are grateful to Franziska Trauzettel for her help in data collection and to Robert Schulz for polishing the English.

Notes

1. The focus of egocentric network analysis lies on the relations of one or more actor/s (ego) (e.g. individuals) and his or her social environment (alters). By this method, ego’s network is established from his or her subjective point of view. Questions touching on ego’s relationships in everyday life (so-called name generators) are posed to identify ego’s contacts (alters) to whom a defined relation is maintained (Gamper, 2015). A famous example of a typical name generator is the Burt generator: ‘With whom did you discuss important personal matters during the last month?’ (Burt, 1984).

2. Impressions of this visualization software may be obtained from www.vennmaker.com and youtube.com.

3. Based on Bearman’s and Parigi‘s (2004) critique of the Burt-generator (see footnote 1) and following the objective of conducting a student relationship survey, we replaced ‘important matters’ by ‘important persons’ and primed respondents with both kin and non-kin network members (cf. Von der Lippe & Gaede, 2013).

4. Compared to network pictures, network maps are somewhat pre-structured by certain coordinates (e.g. concentric circles, sectors; see Gamper et al., 2012). Network pictures typically do not have any specifications (Straus, 2002).

5. Overlap between the resulting samples of studies 1 and 2 was only five students due to different study plans of the courses.

6. Limiting the Likert-items in a questionnaire design to three points only (in order to be comparable to the visual design) may be considered critical due to an unjustified loss of information in the data (Kristoffersen, 2015). Conversely, applying 5-point-scales for arrows or circles in a visual equivalent may also be deemed problematic, as we would criticize that we overly burden respondents asking them to keep oriented in the resulting hairball of network structure – validity and reliability concerns would be legitimately raised, then. We further elaborate on this concern in the general discussion section.

7. All consecutive references to columns refer to Table 1.

8. The transposition rule of Likert-scale points into the metric of VennMaker circles was 5 = 3, 4 = 2.5, 3 = 2; 2 = 1.5, and 1 = 1. We elaborate on potential consequences of this approach in the general discussion section.

9. Defined as the number of reported alter–alter ties divided by the total number of theoretically possible items. We exclude the Ego-ties here for better comparability.

10. The transposition rule of Likert-scale points into the metric of VennMaker arrows was 5 = 3, 4 = 2.3, 3 = 1.7, 2 = 1, and 1 = 0. Again, see the discussion section.

11. All of the interview quotations have been translated from German to English by the authors.

12. Gestalt: this can be described as a kind of picture in her head. Here she thought of her network and how it looks like in her opinion. This is the foundation of her drawings.