Team Trust & Social Presence
5.3 The Study
5.4.1 The Data
The study gained a total of 821 respondents, excluding erroneous data entries such as flatlines, excessive missing data, etc. Table 5.7 shows the number of respondents across the various game communities approached to take part in this study.
829th.org/forums
Community Forum Respondents Mount & Blade 238
War Thunder 169
Dota 2 91
Chivalry 78
Natural Selection 2 78
Arma 77
CS:GO 47
29th ID 43
Total 821
Table 5.7: Respondent numbers
Table 5.8 shows the demographic information of the respondents, while Figure 5.1 shows the range of ages in each game community data set. While gender and age were not being considered as meaningful variables in this study, it is interesting to see the demographic information and get a picture of the population who took part in the study.
Community Forum Average Age (sd) Male/Female Mount & Blade 20 (4.1) 236/2
War Thunder 24 (7.9) 165/2
Dota 2 20 (4.0) 85/6
Chivalry 23 (6.3) 76/1
Natural Selection 2 24 (5.3) 74/4
Arma 24 (7.4) 77/0
CS:GO 22 (4.7) 44/1
29th I.D. 21 (5.1) 43/0
Total 22 (6.0) 98% Male
Table 5.8: Respondent demographic information across games, age (standard deviation), and ratio of male to female respondents. (NA answers excluded)
Arma Chivalry CS:GO Dota 2 Mount & Blade NS2 29th ID War Thunder 10
20 30 40 50
Respondent Ages
Age
Figure 5.1: Ages of respondents by game
Table 5.9 shows information about the context of play across the various game communities, including average team sizes and proportion of respondents basing their experiences of public (pub) play rather than any sort of organized game. Table 5.9 shows that most game experience in this study is based upon public play, with only the Arma community and the 29th I.D. clan basing their experiences on predominantly organised play.
Community Forum Average Team Size Percentage Pub Play
29th ID 32 32.6%
Mount & Blade 22 74.9%
Arma 16 41.6%
Chivalry 11 84.8%
War Thunder 10 95.2%
Natural Selection 2 9 92.3%
CS:GO 5 91.5%
Dota 2 5 94.6%
All Data 14 79.5%
Table 5.9: Context of Play, (NA answers excluded)
Table 5.10 shows the number of respondents in relation to the number of views the call for participants forum thread received, in addition to the daily peak players of each game around which the communities were based (not including Chivalry and 29th I.D. due to missing forum view data). The results shows that on average for every eight people that viewed a forum thread, one person proceeded to take part in the survey. This is close to the one in ten of previous studies, with the increase likely due to the offer of the game prize draw.
Community Forum Respondents Thread Views Response/View Rate Peak Players
Mount & Blade 239 2427 1/10 8000
War Thunder 169 884 1/5.2 9000
Dota 2 92 717 1/7.8 600,000
Arma 3 78 702 1/9 21,000
Natural Selection 2 78 534 1/6.8 1000
CS:GO 47 357 1/7.6 110,000
Total 703 5621 1/8
Table 5.10: Responses and Thread Views of all community forums. Peak daily Player counts are rounded down to the nearest thousand, are based upon numbers from the Steam statistics site (store.steampowered.com/stats/) from March 2014. Chivalry and 29th I.D. not included due to miss-ing forum view data.
5.4.2 Factor Analysis
Principal component analysis was conducted on the data set to explore if the sub-scales were measuring factors as expected. Conducting factor analysis on a validated questionnaire is advocated by Kline [2014]
when using multiple scales. In addition, the large data-set provided an opportunity to re-evaluate the structure of the questionnaire following the inconclusive results of the cooperative module in the previous Chapter. The factor analysis was conducted following Kline [2000] with the oblique rotation method, direct oblimin. The scales used in this study were the CCPIG and the Langfred [2004] trust scale. The CCPIG contains competitive and cooperative sections which aim to measure separate (but related) concepts, and thus the items in these sections are expected to load onto distinct factors. The competitive section contains two modules which were also expected to be separate factors, and a cooperative section which previous analysis had established as one large factor. It was expected the the trust scale would measure a fourth factor. While the factor analysis confirmed the competitive/cooperative split, and the separate competitive modules (Table 5.11), the cooperative and trust modules did not split as expected. In the following tables showing the results of the PCA, loadings of over 0.4 are considered noteworthy and are highlighted in blue to denote factor loadings.
Scales Item Comp. 1 Comp. 2 Comp. 3 Comp. 4
Table 5.11: PCA of all scales, 4 factor split, over 0.4 highlighted
When testing only the CCPIG data (Table 5.12), there is a clear split between the cooperative and com-petitive sections, showing that the questionnaire is broadly measuring the factors as it should.
Item Comp. 1 Comp. 2 Comp. 3
Table 5.12: PCA of CCPIG, over 0.4 highlighted
When taken on its own, the competitive section splits clearly into its respective modules (Table 5.13).
Item Comp. 1 Comp. 2
Table 5.13: PCA of CCPIG competitive modules, over 0.4 highlighted
Table 5.11 shows that the cooperative section contained a high degree of cross-loading, suggesting it was measuring one single factor as in previous studies. However if we take a more pragmatic approach we can see that the majority of cross-loading items load far more strongly on one of the two factors (Table 5.14).
The following table highlights loadings greater than 0.6, rather than 0.4.
Item Coop. 1 Coop. 2
Table 5.14: PCA of CCPIG cooperative section, over 0.6 highlighted
These strongly loading items can be split into two distinct modules with coherent themes. Items in the first cooperative factor, which we shall call Module 2.1, predominately refer to the cohesion and effectiveness of the team, while items in the second factor (Module 2.2) refer to the interplay and involvement between the player and the team. In broad terms Module 2.1 appears to be measuring perceived team cohesion, while Module 2.2 seems to be measuring the player’s involvement or investment in the team.
Cooperative Module 2.1 Cooperative Module 2.2
Perceived Team Cohesion Team Involvement
I felt like I was part of a team I was aware of my team
I felt a social connection to my team-mates/camaraderie I acted with my team-mates in mind team-mates/camaraderie
I felt my team-mates were looking out for me I considered my team-mates possible plans/thoughts
I felt the team helped me I felt I contributed to the team My team-mates played a significant role in I felt my actions made a difference to
my game experience my team-mates
My team communicated well The actions of my team-mates affected my thoughts and actions
The team had a mutual understanding My actions were determined by the objectives of the team
I put the performance of the team over my I wanted my team to value me personal performance
Being part of a team motivated me I felt responsible for achieving the objectives of the team
I felt my team was committed to working together I made an effort to work with my team-mates My team-mates were useful I did not want my team to think I had let them down
had let them down I felt my team shared a common overall aim
I felt my team shared common short term goals It was as much about the team as about my own game
Table 5.15: Cooperative Social Presence Items split in to Modules 2.1 & 2.2
When a PCA was performed on the data from both the cooperative social presence and team trust measures (Table 5.16), the two scales appeared to be measuring a single factor. The results show a large amount of cross loading across the cooperative scale, and that the majority of the team trust scale strongly loaded onto the first factor. The preliminary factor analysis of all the scales (Table 5.11) suggested that team trust and cooperative social presence shared common factors, while Table 5.16 confirmed there is no clear split between the two scales. While it was expected that there was some conceptual crossover between team trust and cooperative social presence the results of the PCA suggest a strong conceptual link between the two scales.
Item Comp. 1 Comp. 2
The main scales used in this study were the Langfred [2004] four item trust scale, and the CCPIG, consisting of a two factor competitive social presence section and newly established two factor cooperative social presence section. The following statistics show that each scale had good KMO and Cronbach α scores.
The only scale which failed to achieve a strong Cronbach’s α was the competitive Module 1.1, however as a whole the competitive section had strong internal consistency. Therefore, the data seems to suggest that the reliability of the scales was adequate, and the factor analysis showed that the CCPIG was broadly working as expected.
To summarize the scales/modules used in the data analysis:
Competitive & Cooperative Presence in Gaming (CCPIG) Questionnaire
Section 1: Competitive Social Presence - This section of the CCPIG measures the level of social presence a respondent felt towards their opponent.
Module 1.1: Awareness - Measures how aware a respondent was of their opponent.
Module 1.2: Engagement - Measures how challenging and engaging a respondent felt
their
opponents were.
Section 2: Cooperative Social Presence - Section 2 measures the level of social presence a respondent
felt towards their team-mates.
Module 2.1: Cohesion - Measures a how cohesive and effective a respondent felt their team was.
Module 2.2: Involvement - Measures how involved and invested a respondent felt they were in
their team.
Trust - The Langfred [2004] scale measures the level of trust a players perceived existed in their team.
A full list of items can be found in the Appendix.
Cronbach’s α for the ’Competitive’ data-set Items: 14
Sample units: 821 α: 0.807 KMO: 0.855
Cronbach’s α for the ’Competitive 1.1 Awareness Module’ data-set Items: 6
Sample units: 821 α: 0.675 KMO: 0.767
Cronbach’s α for the ’Competitive 1.2 Engagement Module’ data-set Items: 8
Sample units: 821 α: 0.773 KMO: 0.841
Cronbach’s α for the ’Cooperative’ data-set Items: 25
Sample units: 821 α: 0.942 KMO: 0.952
Cronbach’s α for the ’Cooperative 2.1 Cohesion Module’ data-set Items: 14
Sample units: 821 α: 0.935 KMO: 0.951
Cronbach’s α for the ’Cooperative 2.2 Involvement Module’ data-set Items: 11
Sample units: 821 α: 0.841 KMO: 0.856
Cronbach’s α for the ’Trust’ data-set Items: 4
Sample units: 821 α: 0.813 KMO: 0.766