Correlations
Pearson's correlations indicated that the psychological measures were positively correlated, ranging from r = .49 to .78 (Table 2, shaded area). The BMI did not correlate significantly with any FFFT or PWB measures. Furthermore, the correlations between FFFT components were positive, and apart from one (FL4 and FL2), all were statistically significant (Table 2). Several PWB measures correlated statistically significantly with FFFT measures (Table 2, in bolded values). In the case of FL5, the negative correlations indicate a positive connection (shorter time means better coordination).
Table 2 Correlations between physical functionality tests, BMI, and psychological measures
Ordinary Least Squares (OLS) Regressions
As shown in Table 3, only FL6 was significant in only two models, despite all predictors having a high correlation with PWB variables. So, due to multicollinearity, we cannot determine the importance of the predictors based on OLS regression results.
Table 3 Ordinary Least Squares Regressions – multicollinearity present (n = 39)
|
Resilience
|
Mental well-being
|
Optimism
|
Life satisfaction
|
Happiness
|
Predictors
|
b
|
p
|
b
|
p
|
b
|
p
|
b
|
p
|
b
|
p
|
Intercept
|
1.93
|
.070
|
3.41
|
.011*
|
0.95
|
.439
|
2.37
|
.270
|
4.43
|
.056
|
FL1
|
0.02
|
.484
|
0.03
|
.287
|
0.02
|
.508
|
0.03
|
.525
|
0.03
|
.594
|
FL2
|
-0.01
|
.766
|
-0.01
|
.728
|
-0.03
|
.350
|
0.02
|
.691
|
-0.02
|
.674
|
FL3
|
-0.00
|
.629
|
0.01
|
.537
|
0.00
|
.895
|
-0.02
|
.297
|
0.00
|
.981
|
FL4
|
0.02
|
.351
|
0.04
|
.118
|
0.03
|
.200
|
0.06
|
.189
|
0.06
|
.153
|
FL5
|
-0.02
|
.604
|
0.02
|
.635
|
0.05
|
.144
|
0.01
|
.902
|
-0.01
|
.881
|
FL6
|
0.01
|
.110
|
0.01
|
.197
|
0.02
|
.005*
|
0.02
|
.021*
|
0.01
|
.320
|
BMI
|
0.01
|
.783
|
0.02
|
.437
|
0.03
|
.307
|
-0.02
|
.639
|
-0.02
|
.720
|
R2 / R2 adjusted
|
0.460 / 0.338
|
0.391 / 0.253
|
0.369 / 0.227
|
0.517 / 0.408
|
0.359 / 0.215
|
Note. FL1 to FL6 = Fullerton test forms; BMI = Body mass index; b = unstandardized regression coefficient; p = p-value * = statistically significant
Elastic Net Regressions and LMG Relative Importance
Resilience
Concerning resilience, the LMG method and elastic net model indicated that the FL6, FL5, FL1, and FL4 were the most important (Figure 1, a, c), while FL3 and FL2, and BMI were the least significant predictors. As such, FL3, FL2, and BMI were shrunk to zero in the elastic net model. FL5 was negatively related to the outcome since a shorter time indicates better performance. Optimal values for lambda and alpha hyperparameters were 0.275 and 0.125, respectively (Figure 1).
Mental well-being
As for MWB, FL4 and FL6 appeared to be the most important, while again, FL2 and BMI were the least significant predictors. FL3 was not shrunken to zero in the elastic net model, having roughly equal magnitude as FL6, although the LMG metric indicated FL6 had much greater importance than FL3. Finally, FL1 and FL5 could be equally important predictors. The best lambda value for this model was 0.789, and the alpha was set to 0.1 (Figure 2).
Optimism
Regarding Optimism, FL6 was shown to be the most important predictor, and its contribution was notably larger than the rest, followed by FL4 and FL5. Again, FL2, FL3, and BMI were the least important. However, the elastic net shrunk all but FL6 and FL4 predictors to zero. Alpha and lambda hyperparameters for this model were .216 and 0.258 (Figure 3).
Satisfaction with life (SWL)
For SWL, FL6 was again the most important, followed by FL5 and FL1. BMI, FL2 and FL3 were, again, the least relevant predictors. Surprisingly, FL1 played a more important role than the other well-being outcomes (with the exception of resilience), and FL5 was reduced to zero in the elastic net regression model. Optimal values for alpha and lambda were 0.729 and .228 (Figure 4).
Happiness
FL4 seemed the most important for Happiness, followed by FL6 and FL5. Once again, FL2, BMI, and FL3 appeared the least important in predicting this outcome. Only FL2 was shrunk to zero in the elastic net. For happiness, the chosen alpha value was 0.1, and the optimal value for lambda was 1.223 (Figure 5).