Statistical Methods Used in Analysis of Data

    All assumptions of the statistical methods used were checked and met, with normality checks and Cronbach’s Alpha calculated where applicable. Statistical tests used were Pearson Product-Moment correlations, Multiple Regression, ANOVAs and t-tests.

PVP means and relationships with gender and age.

The mean score of the PVP was 4.17, with SD = 2.19. Although the discrepancy in sample size across gender is inhibitive of detailed comparisons such as running two separate multiple regressions, it was found that the PVP total score and hours of play per week were not significantly different across genders according to independent groups t-tests. Age was not related to PVP scores.

PVP item endorsement.

The percentage of the endorsement for the PVP Items, in order of endorsement, is presented in Table 1 below. Although all items are summed to acquire the total score used in analyses, thereby giving each item an equal weight, the individual item endorsement indicates which video game playing problems are most common. The item pertaining to withdrawal was the least endorsed, although 45.7% of the sample played games more when they felt bad or had problems. The item with many different forms of loss of control showed the highest endorsement, but the question is unable to distinguish which exact problem (from six) was endorsed.

Table 1

 

Item Endorsement of the PVP in Order of Endorsement

 

 

PVP validity.

Similar to the normative data the PVP is based on, a significant positive correlation was found between PVP total score and average hours of game play per week (r = .317, N = 621, p < .001), and with average hours of game play per day (r = .298, N = 621, p < .001). However, the magnitude of these relationships are smaller than those reported by the authors (Salguero & Moran, 2002), between the PVP total score and play frequency (r = .64, N = 223, p < .001), duration (r = .52, N = 223, p < .001) and longest play time per session (r = .56, N = 223, p < .001). The internal consistency was also found to be lower than previous studies (α = .661).

RSES and SSI subscale validity.

Chronbach’s Alpha scores were also calculated for the RSES (α = .90), as well as all six SSI subscales; including EE (α = .71), ES (α = .8), EC (α = .74), SE (α = .9), SS (α = .84) and SC (α = .86).
 
Relationships Between Problem Video Game Play, Social Skills and Self Esteem

To investigate potential relationship between problem video game play, social skills and self esteem, Pearsons Product-Moment correlations were generated between the PVP total score (all nine items summed), and SSI subscales and the RSES total score. No relationships were found between the PVP and non-verbal subscales EE (r = -.044, N = 621, p = .27), ES (r = -.045, N = 621, p = .26) and EC (r = -.057, N = 621, p = .15). However small significant relationships were found between the social subscales SE (r = -.134, N = 621, p = .001), SS (r = .224, N = 621, p < .001) and SC (r = -.217, N = 621, p < .001). It is noteworthy that significant relationships were only present in the verbal subscales but not the nonverbal. It is apparent that as video game playing problems increase so does SS, whereas SC and SE decrease. Duplicating past studies, the strongest relationship that emerged was between the PVP and RSES (r = -.234, N = 621, p < .001).

Relationships Between Problem Video Game Play and Dimensions of Social Skills

Certain SSI subscales can also be summed to provide indications of important constructs within the SSI, including Total Social (SE + SS + SC) and Total Emotional (EE + ES + EC); as well as Total Expressivity (SE + EE), Total Sensitivity (SS + ES) and Total Control (EC + SC). The subscales can be summed to provide an overall SSI score, however caution must be taken when interpreting these figures as a balance of the subscales is emphasised in interpreting the SSI summed scores. A higher overall score, without an interpretation of balance between the subscales, does not necessarily indicate a socially skilled person. The SSI provides a formula to measure this balance . As contemporary studies have related this balance score to severity of pathological symptoms , this will also be included in analysis.
Pearsons correlations were run between the PVP and the aforementioned variables. A very small inverse relationship was present between the overall score and the PVP (r = -.084, N = 621, p = .037), as well as a small positive relationship between Total Sensitivity (r = .130, N = 621, p < .001), and small inverse relationships between Total Expressivity (r = -.108, N = 621, p = .007) and Total Control (r = -.187, N = 621, p < .001). Relationships between the PVP and Total Social (r = -.077, N = 621, p = .056) and Total Emotional (r = -.075, N = 621, p = .06) approached significance. The balance score was not significantly related to PVP, hours of play per week or self esteem. Due to the emphasis on balance when scoring the SSI, and these relationships being smaller in magnitude than those between the PVP total score and the individual subscales (as well as being partial products of them), an interpretation of the individual subscales is a more useful indicator of the relationships between the PVP and the SSI.

Predicting Problem Video Game Play with Social Skills and Self Esteem

Multiple regression was employed to determine if a measure of social skills and self esteem could predict the magnitude of reported problematic play. Using the enter method, prediction of the PVP total score was attempted using the predictor variables of the six SSI Subscales, the RSES, sex and the average hours of play per week. Because sex is a dichotomous variable, it was entered into the model as a ‘dummy’ variable, with females coded as zero and males as one . Using the enter method, a significant model emerged: F (9, 611) = 15.051, p < .001. However, the model explains a negligible 16.9% of the variance in the PVP total score (adjusted R2 = .169), with the predictor variable of average hours of play per week accounting for the highest prediction (β = .314, p < .001). The two other significant predictor variables were the SSI subscale SS (β = .184, p < .001) and the RSES total score (β = -.103, p = .03). Interestingly, SE and SC showed correlations with PVP scores, but neither emerged as significant predictors in a regression model. Sex approached significance (p = .056). The results of this multiple regression were confirmed using the stepwise method. Table 2 below shows a breakdown of the predictor variables using the enter method.

Table 2

 

Summary of Enter Multiple Regression Analysis Predicting PVP Total Score

 

 

*p<.05. **p<.000

Relationships Between Time Spent Playing Online Games and Self Esteem and Social Skills

To investigate the relationship between time spent playing online games and social skills and self esteem, an additional Pearsons correlation matrix was produced. Participants that reported they play primarily offline games were excluded from the analysis (n = 194), and correlations were run between average hours of game play per week and the SSI subscales and the RSES. Interestingly, no significant relationships emerged, suggesting a more prominent relationship between problems associated with game play and social skills and self esteem, not amount of online game play alone. A multiple regression was run, with average hours of play per day as the criterion variable, and all six SSI subscales and the RSES total score as predictor variables. Using the enter method, an insignificant model was found: F(7, 419) = 1.62, N = 427, p < .12, confirming the absence of significant relationships between time spent playing online games and self esteem and social skills.

Genre Comparison

To determine if differences were present between the 5 game genres on PVP total scores, average hours of play per week, all SSI subscales, RSES total score and age, one way between-subjects ANOVAs were conducted. Those that selected “not applicable” for genre were excluded from analysis (n = 8). No significant differences were found between genres on all measures except for PVP total scores (F(4, 607) = 4.054, p = .003, partial η2 = .02), hours of game play per week (F(4, 607) = 32.843, p < .001, partial η2 = .17), and age (F(4, 611) = 6.394, p < .001, partial η2 = .04). However, for hours of play per week, Levene’s test for Homogeneity of Variance was significant, indicating a violation of the assumptions of normality for this ANOVA, and post-hoc comparisons of group means must be interpreted with caution.

     Post-hoc hours of play per week analysis.

MMORPG players showed the highest mean average hours of play per week (M = 34.16, SD = 19.32), which was significantly higher than the next highest mean of action players (M = 21.32, SD = 14.26), according to a Bonferonni post-hoc test (p < .001). As homogeneity was not met for this ANOVA, an independent groups t-test was conducted comparing MMORPG to action players to confirm the result. As Levene’s test for equality of variance was significant, the result for equal variances not assumed was used. A significant difference was found (t = -7.725, df = 301.542, p < .001).

Post-hoc PVP mean scores analysis.

Similarly, MMORPG players showed the highest mean PVP total score (M = 4.69, SD = 2.29), however, using Bonferonni post-hoc tests, it was not significantly different from the next highest mean of strategy (M = 4.19, SD = 2.07), but was significantly different from action (p < .02), the next highest mean score (M = 4.06, SD = 2.13).

Post-hoc age analysis.

MMORPG players had the highest mean age (M = 25.76, SD = 7.31), which was significantly higher than action players (M = 22.72, SD = 6.10), according to Bonferonni post-hoc (p < .001).

Age, and Age First Played Electronic Games

Using Pearson correlations, age was not related to PVP or hours of play per week, but it was minimally related to subscales EC (r = -.110, N = 621, p = .006), SS (r = -.094, N = 621, p < .02) and SC (r = .119, N = 621, p = .003). Confirming past studies, a small positive correlation between age and RSES total scores was found (r = .173, N = 621, p < .001).
    Age at which first play electronic games was not related to any SSI subscales or measures of play time, but was minimally related to RSES scores (r = .082, N = 621, p < .04).