The SEM was performed to confirm the PCDEQ2 (
1). In this study, a seven-factor model was tested. Fit indices, including RMSEA = 0.056, TLI and CFI > 0.9, PCFI = .751, and CMIN/DF = 2.15 were obtained, which were all acceptable. Standardized factor loadings (from 0.4 to 0.74) were significant at the 0.001 level (
Table 2). Data showed adequate fit to the 7-factor model: χ
2 (3737) = 9522.86, CFI = 0.839, TLI = 0.818, RMSEA = 0.054, 90% CI (0.053, 0.056), PCFI = 0.682, CMIN/DF = 2.55, SRMR = 0.079. Evaluation of factor loadings showed that factor loadings of items 21, 50, 58, and 83 were smaller than 0.3, so these items were eliminated. Inspection of modification indices indicated that error terms for items 4 and 7 (χ
2 = 59.44), 8 and 9 (χ
2 = 118.88), 73 and 71 (χ
2 = 84.34), as well as items 51 and 52 (χ
2 = 104.35) had relatively large modification indices compared to the others. The measurement model was thus re-specified by having the above-mentioned pairs of error terms as correlated. The model fit of the specified model was improved: χ
2 (3391) = 8455.593, CFI = 0.918, TLI = 0.91, RMSEA = 0.053, 90% CI (0.052, 0.055), PCFI = 0.712, CMIN/DF = 2.49, and SRMR = 0.078. Internal consistency reliability for the whole questionnaire was obtained using Cronbach's alpha (α = 0.89). All factors revealed internal reliability above the minimum recommendations of 0.70, ranging from 0.703 to 0.924.
4.1. Competition Level
Next, we wished to compare the discriminant capability of the Persian version to those obtained with the original. The total number of the high-level and low-level athletes identified among the participants was 69 and 459, respectively. Consistent with previous studies (
24,
25), those athletes who competed in various sports at national and international levels were defined as the high competitive level athletes, and those who competed at a recreational or club level were defined as low competitive level athletes. Based on previous research suggesting that certain athlete characteristics, such as competitive level, influence PCDE (
12,
24-
28), it was hypothesized that athletes at higher levels of competition have different psychological characteristics than athletes at lower levels of competition. Before the ANOVA analysis, since there were 69 athletes in the high levels of competition, the sample was randomly removed from low levels of competition to equal the number in the groups, and the analyses were performed with 138 samples (69 people in each group). Our findings supported this hypothesis.
As with the UK applications of the PCDEQ2, high-level and low-level athletes have different PCDE profiles (
Table 3). A Mahalanobis distance was obtained, 58.19, which is less than the critical value of 61.3, indicating multivariate normality (Tabachnick and Fidell, 2013). We found a difference in PCDEQ between the groups high-level and low-level, [F (7, 130) = 4.55, P < 0.001, Wilks Lambda = 0.803, partial eta squared = 0.197].
Table 2 shows descriptive statistics of PCDEQ in two groups, effect sizes, and significance levels of differences between groups. Furthermore, we found a significant discriminant function of the PCDEQ (Wilks Lambda = 0.803, χ
2 = 29.07, P < 0.001), with a canonical correlation of 0.444. The PCDEQ was able to correctly predict 66.7% (46 out of 69) of the high-level competition group members and 73.9% (51 out of 69) of the low-level competition group members; in total, 70.3% of the 138 participants could be correctly classified.
| Factors | Low-level | High-level | Effect Size | P-Value |
|---|
| Adverse response to failure | 3.27 ± 0.83 | 2.93 ± 0.82 | 0.412 | < 0.05 |
| Imagery and active preparation | 4.94 ± 0.65 | 4.5 ± 0.91 | 0.563 | < 0.01 |
| Self-directed control and management | 4.06 ± 0.89 | 4.34 ± 0.59 | 0.371 | < 0.05 |
| Perfectionistic tendencies | 3.57 ± 0.96 | 3.08 ± 0.72 | 0.582 | < 0.001 |
| Seeking and using social support | 4.17 ± 0.93 | 4.64 ± 0.73 | 0.55 | < 0.001 |
| Active coping | 4.22 ± 0.99 | 4.54 ± 0.505 | 0.407 | < 0.01 |
| Clinical indicators | 3.21 ± 0.99 | 2.73 ± 0.78 | 0.538 | < 0.01 |
a Values are expressed as mean ± SD.
b Responses were provided on a 6-point Likert scale from 1 (“very unlike me”) to 6 (“very like me”).
To examine gender, sport type, and age invariance, first the unrestricted models were examined, and then the fit indices of the models with factor loading restrictions were compared with the unrestricted model in
Table 4. The ∆χ
2 values, which were calculated with the aim of the chi-square test of the constrained and unrestricted models, showed that the factor loadings (P = 0.234, ∆χ
2 = 434.24) among girls and boys, the factor loadings in individual and team sports (P = 0.326, ∆χ
2 = 355.79), and the factor loadings in age groups (P = 0.315, ∆χ
2 = 370.445) are equal. Accordingly, the invariance of gender, sport type, and age in the psychological characteristics of the developing sports elite is confirmed.
| Factors | Individual | Team | Effect Size | P-Value |
|---|
| Adverse response to failure | 3.25 ± 0.94 | 3.14 ± 0.92 | 0.12 | < 0.05 |
| Imagery and active preparation | 4.37 ± 0.88 | 4.9 ± 0.68 | 0.674 | < 0.001 |
| Self-directed control and management | 4.15 ± 0.67 | 4.37 ± 0.66 | 0.330 | < 0.001 |
| Perfectionistic tendencies | 3.49 ± 0.88 | 3.43 ± 0.85 | 0.069 | > 0.05 |
| Seeking and using social support | 4.27 ± 0.89 | 4.5 ± 0.71 | 0.285 | < 0.01 |
| Active coping | 4.27 ± 0.68 | 4.52 ± 0.61 | 0.387 | < 0.001 |
| Clinical indicators | 3.3 ± 0.81 | 2.89 ± 0.77 | 0.519 | < 0.001 |
a Values are expressed as mean ± SD.
b Responses were provided on a 6-point Likert scale from 1 (“very unlike me”) to 6 (“very like me”).
4.2. Other Mediating Factors: Gender and Sport Type
The differences between males and females in PCDEQ are presented in
Table 4, and the differences between athletes in team and individual sports are presented in
Table 5. As can be seen in
Table 1, the number of sample people in the groups of male and female, as well as team and individual sports, are different, so the sample was randomly removed from the groups to equal the sample size in all groups (205 people in team and individual sports in each group and in women and men of each group, 219 people).
| Factors | Male | Female | Effect Size | P-Value |
|---|
| Adverse response to failure | 3.23 ± 0.88 | 2.97 ± 0.83 | 0.304 | < 0.001 |
| Imagery and active preparation | 4.64 ± 0.81 | 4.95 ± 0.65 | 0.422 | < 0.001 |
| Self-directed control and management | 4.23 ± 0.68 | 4.44 ± 0.58 | 0.332 | < 0.001 |
| Perfectionistic tendencies | 3.45 ± 0.82 | 3.53 ± 0.902 | 0.093 | > 0.05 |
| Seeking and using social support | 4.26 ± 0.82 | 4.51 ± 0.74 | 0.320 | < 0.001 |
| Active coping | 4.22 ± 0.69 | 4.55 ± 0.56 | 0.525 | < 0.001 |
| Clinical indicators | 3.11 ± 0.86 | 2.8 ± 0.73 | 0.389 | < 0.001 |
a Values are expressed as mean ± SD.
b Responses were provided on a 6-point Likert scale from 1 (“very unlike me”) to 6 (“very like me”).
We found a difference in PCDEQ between team and individual sports, [F (7, 402) = 9.96, P < 0.001, Wilks Lambda = 0.852, partial eta squared = 0.148].
Table 3 shows descriptive statistics of PCDEQ in two groups, significance levels of differences between groups, and effect sizes. The results show that the difference between athletes in team and individual sports is obtained in significant dimensions except in dimensions adverse response to failure and perfectionistic tendencies. We also found a difference in PCDEQ between males and females, [F (7, 430) = 7.902, P < 0.001, Wilks Lambda = 0.886, partial eta squared = 0.114]. Descriptive statistics of PCDEQ in two groups, effect sizes, and significance levels of differences between groups are presented in
Table 5.