A common test for bias
To minimize bias between the two surveys towards a single dimension, a procedure strength and Harman's one-way test were used to evaluate the questionnaire format, which was disseminated online using Questionnaire Star and had bolded question stems and titles. There was 1 eigenroot component, and the eigenvalues of the major factors were T1=19.323% and T2=21.542%. According to the principle of the common method bias test, the primary features accounted for less than 40% of the variance, indicating that the common method bias of the survey test is acceptable. [10]
Analysis of the descriptive characteristics and correlations between mental health, physical exercise, and risk perception
Independent samples The t-test was used to assess the gender differences in each variable (see Table 1), and the results revealed stable gender differences in mental health (P<0.001), risk perception (P<0.001), and physical activity (P<0.001) over time for both T1 and T2. In addition to the comparison of means in Table 2, it was discovered that men reported greater mental health scores than women and that women had higher scores on two repeated tests. The risk perception levels of women were higher than those of males. Mental health and risk perception had effect sizes of 0.159 (d=0.262) and 0.179 (d=0.278) for men and 0.255 (d=0.416) and 0.195 (d=0.402) for women, respectively.
Table 1 Gender-independent sample t-test
Variables
|
HV-test
|
Levene-test
|
T-test
|
F
|
P
|
T
|
df
|
P
|
95%CI
|
LLCI
|
ULI
|
T1 mental health
|
variance chi-squared
|
2.593
|
0.109
|
4.265
|
344
|
0.0012)
|
0.676
|
0.886
|
T2 mental health
|
variance chi-squared
|
0.422
|
0.969
|
2.512
|
344
|
0.0042)
|
0.518
|
0.883
|
T1 physical activity
|
variance chi-squared
|
0.438
|
0.509
|
-0.653
|
344
|
0.292
|
-0.348
|
0.693
|
T22 physical activity
|
variance chi-squared
|
0.231
|
0.969
|
-1.056
|
344
|
0.515
|
-0.394
|
1.130
|
T1 risk perception
|
variance chi-squared
|
0.910
|
0.341
|
2.513
|
344
|
0.0011)
|
0.116
|
0.885
|
T2 risk perception
|
variance chi-squared
|
0.584
|
0.445
|
2.868
|
344
|
0.0011)
|
0.355
|
0.916
|
1) P<0.001;2)P<0.01
To exclude interference from demographic variables, a partial correlation analysis was undertaken for the three variables of mental health, physical activity, and risk perception, and gender was added for comparison to obtain more precise results. (See Table 2.) T1 mental health and T2 mental health were strongly positively associated (p<0.001), as were T1 risk perception and T2 risk perception. The positive correlation between T1 mental health, T1 physical exercise, and T1 risk perception in the first questionnaire survey (p<0.001) and T2 mental health, T2 physical exercise, and T2 risk perception in the second questionnaire survey (p<0.001) indicated the consistency of the recovered data and the synchronization of the correlations.
Table 2 Results of descriptive statistics and partial correlation analysis for each variable
Variables
|
T1 mental health
|
T2 mental health
|
T1 physical activity
|
T2 physical activity
|
T1
risk perception
|
T2
risk perception
|
T1 mental health
|
1
|
|
|
|
|
|
T2 mental health
|
0.6721)
|
1
|
|
|
|
|
T1 physical activity
|
0.4321)
|
0.4231)
|
1
|
|
|
|
T2 physical activity
|
0.4361)
|
0.3631)
|
0.5891)
|
1
|
|
|
T1 risk perception
|
0.5521)
|
0.3321)
|
0.2141)
|
0.3741)
|
1
|
|
T2 risk perception
|
0.3071)
|
0.3511)
|
0.1781)
|
0.5451)
|
0.2741)
|
1
|
General M±SD
|
2.812±0.189
|
1.534±0.165
|
2.375±0.124
|
3.684±0.21
|
3.707±0.119
|
1.810±0.152
|
Male M±SD
|
2.815±0.242
|
1.548±0.199
|
2.381±0.150
|
3.703±0.242
|
3.694±0.151
|
1.819±0.191
|
Female M±SD
|
2.804±0.291
|
1.547±0.304
|
2.363±0.223
|
3.657±0.403
|
3.733±0.194
|
1.891±0.249
|
1) P<0.001;2)P<0.01
Analysis of risk perception, physical activity, and adolescent mental health across periods
The cross-lag analysis of risk perception, physical exercise, and mental health of adolescents was conducted by pre-correlation analysis with bias scores according to the packing and dimensionality reduction technique, and the data model was constructed using AMOS25.0 statistical software (Figure 1). In the cluster analysis, it was necessary to restrict various parameters to find the most appropriate model, by comparing the pre-defined model, the path coefficient equation, and the model with the fewest number of parameters. Comparing the fitness of six models—the preset model, the path coefficient equality model, the covariance equality model, and the variance equality model—led to the selection of the measurement error model as the analytical model. The data were good for model fit results, X²/df=1.713 (p<0.01); absolute fit index RMSEA=0.082; relative fit index CFI=0.911, GFI=0.931, NF1=0.935, IF1=0.959, TLI=0.956, and risk perception T1, physical exercise T2, and mental health T2,3 The variables were created as distinct models of mediating effects, and the Bootstrap method was used to examine the mediating effect of physical exercise T2 between risk perception T1 and mental health T2. The foundation of the Bootstrap approach was to investigate the correlation of a*b. On the one hand, the Sobel test is conducted, which has high data needs, a big sample size, and a normal distribution, resulting in low testing efficiency. The sampling test procedure for the initial sample is contrasted with this. The Bootstrap sampling approach is a more prevalent test due to its efficiency, and there are no limits on the distribution pattern of the sampling for the mediating effect. The bootstrap sampling method is based on repeated sampling of the initial sample, and the significance of the coefficient of the mediating effect is evaluated using a 95% confidence range (CI). [11] Hayes (2009) suggested that the initial sample is sampled up to a thousand times for the Bootstrap mediation effect test. If the results of the Bootstrap mediated effects test reveal that the Bootstrap test CI does not contain the value 0, the indirect impact will begin to take effect (Chen, R. et al., 2013). In this study, the technique of calculating the mediating effect Bootstrap 95% CI was based on sampling the sample 1000 times for the mediating effect test, and the findings are presented in Table 3. The point estimate of the direct effect of risk perception T1 → physical exercise T2 → mental health T2 was 1.010, the mediating effect was 0.012, the Z value was 0.112, and the indirect effect Bootstrap 95% CI from this path did not have a 0 value, indicating a significant mediating effect of physical exercise between risk perception and mental health.
Table 3 Intermediation effect result test
item
|
Total effect(C)
|
a
|
b
|
Intermediary value effect(a*b)
|
Boot SE
|
Z-value
|
P-value
|
95% BootCI
|
Direct effect
|
Test conclusion
|
risk perception T1→physical exercise T2→mental health T2
|
1.022**
|
0.048
|
0.245
|
0.012
|
0.105
|
0.112
|
0.911
|
0.197~0.229
|
1.010**
|
Significant intermediation
|
The results demonstrated that the influence of mental health T1 on physical exercise T2 (β=0.08) was statistically significant, as was the effect of physical exercise T1 on mental health T2 (β=0.33) and the effect of risk perception T1 on mental health T2 (β=0.28) and physical exercise T2 (β=0.19). There was no statistically significant relationship between mental health T1 (β=-0.01) and physical activity T1 (β=-0.27) and risk perception T2 (p>0.05).