P. 1
PhD09_vanderGraaf

PhD09_vanderGraaf

|Views: 67|Likes:

More info:

Published by: Shenja van der Graaf on Dec 10, 2010
Copyright:Attribution Non-commercial

Availability:

Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less

11/09/2011

pdf

text

original

The results of the online survey, interviews and document analysis provided the

basis to analyse, or ‘rez’98

in Second Life jargon, the case study. The objective of the
data analysis process was to create a link between the data and the conceptual
framework. This was done by making sense of the empirical data by relating the
empirical indicators to the theoretical concepts, and by checking the validity of the
findings. The overall analysis used the following main tactics: descriptive and
correlational, and cluster analysis and exploratory factor analysis for the quantitative
data, and a thematic analysis for the qualitative data.
Some basic statistical analysis was conducted to describe and explicate the
survey results. Particularly, the functionalities of the design space were presented using
descriptive and correlation techniques (see Chapter 6). Cluster analysis is a technique to
group individuals based on their responses to several variables, while factor analysis
refers to various techniques that aim to identify groups or clusters of variables based on
people’s responses to those variables.99

For this study, a hierarchical clustering method
that served to identify the number of clusters was employed which was followed by a
non-hierarchical K-Means clustering method. This technique was used to analyse user

98

To rez means to create or make an object appear in Second Life by using the toolkit or by dragging it
from the inventory (Rymaszewski, 2007).

99

Three important uses for the factor analysis are to learn the structure of a set of variables; to develop
and evaluate tests and scales; and, to reduce or simplify complex data sets (Pallant, 2005). Note that
different viewpoints exist about what constitutes the various types of factor analyses. In particular, factor
analysis and principal component analysis have been compared in terms of the employment of
communality estimates. See for an overview Field (2005) and Kline (1994).

99

participation patterns and communication behaviour on the firm-hosted platform guided
by four classification variables (respectively, length in-world visits, design capabilities,
information retrieval and supply; see Chapter 5).
Principal Component Analysis (PCA) was selected in order to examine
correlations between variables. More specifically, the analysis aimed to learn which
linear components could be found in the data and what a specific variable contributed to
that component (Kline, 1994). This technique assumes that the sample used reflects the
population and, therefore, the conclusions are limited to the sample, making it difficult
to generalize the outcome (unless multiple samples are used revealing a similar factor
structure to strengthen external validity which is a point for future research (cf. Field,
2005).100

The following steps were followed to conduct the analysis using SPSS. First,
the suitability of the quantitative data set was assessed by considering the sample size
and ratio of subject items, using the Kaiser-Meyer-Olkin measure of sampling adequacy
and Bartlett’s Test of Sphericity. Second, several factors were extracted by considering
eigenvalues and the scree plot. Third, factors were rotated and interpreted so as to
calculate to what degree variables loaded onto extracted factors.101

Fourth, the reliability
of the scale was tested using Cronbach’s α to check whether it reflected the construct it
measured. The analysis assisted in the detection of underlying structures of information
and communication practices of Second Life users (see Chapter 7).
In order to gain a deeper understanding of user participation in mod
development practices it was important to examine the context of, and latent structures
underlying, the explanations and interpretations of user participation in Second Life.
The interview and documentary data were used to ‘tell the story’ of Second Life
throughout the empirical Chapters 5, 6, and 7. The analysis of these data yielded many
insights into “the ways that people organize and forge connections between events and
the sense they make of those connections” in the context of mod development on the
firm-hosted platform (Bryman, 2004: 412). This form of narrative analysis “link[s]
personal experiences to organizational experiences narratives allow[ing] us to study
organizational reality as constructed and transformed by its actors” (Garcia-Lorenzo,
2004: 47). In Second Life’s case of perpetual development, this method was particularly

100

Although a probability sample is recommended for PCA, the sample used in this study generated
survey results with enough variance. Therefore I was confident to perform this type of analysis.

101

The term factor tends to be used to refer to output, strictly speaking however, the outputs for PCA are

components.

100

helpful because it emphasizes continuity and process, highlighting the users’ sense-
making of their roles as participants with different stakeholders that are embedded in
dynamic relationships involved in the mod phenomenon underpinning the firm-hosted
3D platform (Bryman, 2004; Reissman, 2004). As my analysis focuses on the
connections among the users’ accounts of their experiences, the findings tend to ‘stick’
to the case study, making it difficult to generalize the results (cf. Flick, 2006).
The qualitative data analysis focused on the content or meaning of what was said
(or written) rather than on how it was conveyed. Rather than using, for example, words
and paragraphs as units of analysis as is common in quantitative content analysis, this
study identified common elements reported by research participants across qualitative
data (cf. Haythornthwaite and Gruzd, 2007). This resulted in the organization of the
qualitative data into themes and sub-themes (Fereday and Muir-Cochrane, 2006;
Reissman, 2004). This was achieved by the development of categories and a coding
scheme (see Appendix for Coding Scheme, pp. 271-275). Given my aim to validate and
extend the conceptual framework for this study, initial sets of codes were developed
based on my conceptual framework which is informed by the theories discussed in
chapter 3.102

However, the framework was expanded as themes emerged during the
analytical process which required a new category or sub-theme such as ‘education’ and
‘machinima’ (cf. Bryman, 2004; Kvale, 1996).
Documents and interview transcriptions were coded, checked and rechecked
throughout the coding process for consistency. The analysis was facilitated by the
FileMaker Pro database which functioned as a ‘living document tool’, grouping all the
qualitative evidence into a file structure organized according to the theoretical issues
guiding this research (see Chapter 3). The software assisted in developing a protocol,
analysing and cross-checking themes across content, and developing counts of, for
example, information providers versus commentators (see Chapter 7).103

This
organization of the data helped me to keep track of the context in which observations
had been said or written which otherwise would have appeared to be fragmented
(Bryman, 2004).

This analytical method was made more robust by adopting Yin’s (2003) tactic
towards constructing validity by using multiple data sources, establishing operational

102

These codes correspond to the themes of the survey and interview guides.

103

Several quantitative aspects were investigated using this database (see Chapter 6).

101

measures, and having key informants review drafts. It also provided a basis for
establishing the external validity of the results enabling me to generalize the outcomes
to established theoretical domains and the reliability of the study so that the study could
be repeated by reanalysing the protocol and database developed for this research. The
process of making sense of the results of the data analysis involved identifying
relationships between the themes and coding categories, exploring the properties of user
participation in mod development and the different dimensions of learning
opportunities, and uncovering underlying patterns of firm-user interactions. During the
analysis and writing process, the concepts and theories emerging from the evidence
were compared with, and interpreted in the light of, my initial assumptions and the
results of existing research. This allowed me to check for affirmations and contradictory
findings (Eisenhardt, 1989). The approaches to the quantitative and qualitative data
analysis led to the new theoretical insights which are set out in Chapters 8 and 9.

You're Reading a Free Preview

Download
scribd
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->