Several iterations of multiple correlation, step-wise and hierarchical regression yielded inconclusive results about the antecedents and predictors of depression. In particular, the untreated data with respect to sports participation did not seem to relate to emotional wellbeing, mental health or depression.
The weak link with sports involvement eventually provoked the realization that this predictor variable was being mediated by the other measures of self-worth.
Secondly, the construction of the sports involvement variables was broad enough to accommodate every possible field or PE activity high school students engaged in. This included gym class, about which neither the upbeat and self-possessed nor pessimistic student had any choice but to attend.
Accordingly, one of the most important analytical steps was to:
- Filter out all those who had no sports involvement at all. Comprising as they did half the overall sample, they tended to weaken the force of the analysis by affording no possibility that exercise could obviate depression.
- Since the other definitions of sports involvement also showed weak relationships on first iterations, the decision was made to alter this predictor variable to account for intensity of involvement (this is the compound variable “FitnessxCollectivoxIndividualPresente”). At this level, the variable was also recoded to remove gym class. However, the remaining (binomial) categories of individual and team sports provided enough variability for subsequent correlation and regression analysis.
Finally, the distribution of the raw data on depression scores impelled the researcher to recode the Beck1988 variable, in which just 10.5 percent of the overall sample reported themselves as belonging to the two most intense categories (see Table 6 in the appendices).
On recoding, three of the mediating variables – self-esteem, negative affect and personal self-worth – bore substantive correlations with the recoded depression measure and in the expected directions (see Table 21).
Given the multiplicity of interacting variables, several possibly acting bi-directionally, one goes beyond the fundamental relationships between sports involvement and depression to conduct hierarchical regression that is effectively a test of predicting the criterion variable (depression) from the independent variables: sports involvement, self-perception, self-esteem, satisfaction with life, and negative affect.
One opts for hierarchical multiple regression in this case because we have posed a theoretical construct about the relationships among the predictor variables. At the same time, it was the researcher’s choice from the very beginning how many predictors to measure with the set of study protocols.
Which variables to input in the analysis run? First, one tests the relationship with the recoded “Fitness Presente” variable, and subsequently adds in succession “satisfaction with life,” “personal self-worth,” self-esteem” and “negative affect”.
Examining the R2 values for the first linear regression run on frequency of indulging in sports activities, one obtains:
Table 2.
R Square and R Square Change
Order of Entry
Model 1 : Enter “Fitness Presente Recoded”
Model 1: R square = 0.000009
The predictor “Type of sports x intensity of involvement” alone accounts for less than 0.001% of the variance in “Depression 88 Recoded”.
R2 = 0.000009
Model 2 : Enter “Life Satisfaction” next
Model 2: R square = 0.2034366
The Increase in R square: 0.2034366- 0.000009 = 0.203428
The predictor “Life Satisfaction” accounts for 20.3% of the variance in “Depression 1988 Recoded” after controlling for sports involvement as defined above.
R2 = 0.000009 + 0.203428 = 0.2034366
Model Three: Enter “Personal Self-Worth” third
Model 3: R square = 0.218170
The Increase in R square: 0.218170 – 0.203437 = 0.014734
The predictor “Personal Self-Worth” accounts for just 1.3%% of the variance in “Depression 1988 Recoded”, after sports involvement and “Life Satisfaction” were partialed out from “Personal Self-Worth.”
R2 = 0.000009 + 0.203428 + 0.014734 = 0.218170
Model Four: Enter “Self-Esteem” fourth
Model 4: R square = 0.257280655
The Increase in R square: 0.257280655 – 0.218170 = 0.039111
The predictor “Self-Esteem” accounts for 1.0% of the variance in “Depression 1988 Recoded” after sports involvement, “Life Satisfaction” and “Personal Self-Worth” were partialed out from “Self-Esteem”.
R2 = 0.000009 + 0.203428 + 0.014734 + 0.039111 = 0.257280
Model Five: Enter “Negative Affect” fifth
Model 5: R square = 0.365911071
The Increase in R square: 0.365911071 – 0.257280 = 0.108630
The predictor “Negative Affect” accounts for 11% of the variance in “Depression 1988 Recoded”, after “Frequency Weekly”, “Life Satisfaction”, “Personal Self-Worth” and “Self-Esteem” were partialed out from “Negative Affect”.
R2 = 0.000009 + 0.203428 + 0.014734 + 0.039111 + 0.108630 = 0.365911071
About 37% of the variance in the criterion variable “Depression 1988 Recoded” was explained by the first (.0001%), second (20.3%), third (1.5%), fourth (3.9%) and fifth (11%) predictor variables.
The F change for “Life Satisfaction”, Personal Self-worth,” Self-esteem” and “Negative Affect” are substantial and significance values for all four are p<.01. These suggest that the four independent variables are important predictors of “Depression 1988 Recoded.”
Table 3.
Model1: “Fitness Presente Recoded”
Only 0.01% (.0.001004/10 = 0.000100446) of the variance in the criterion variable “Depression 1988 Recoded” can be accounted for by “Fitness Presente Recoded”. The first model, which includes one predictor variable, resulted in an F ratio of 0.002 with a p >.05.
Model 2: “Life Satisfaction”
Fully 2.3% (22.75/10 = 2.3) of the variance in “Depression 1988 Recoded” can be accounted for by both “Fitness Presente Recoded” and “Life Satisfaction.” The second model, which includes two predictors (X1 and X2), resulted in a high F ratio of 34.35 with a p <.001.
Model 3: Personal Self-Worth
About 2.4% (24.4/10 = 2.4) of the variance in “Depression 1988 Recoded” can be accounted for by all three predictors (X1, X2 and X3). The third model, which includes all three predictors, resulted in a high F ratio of 24.92852118 with a p <.001.
Model 4: Self-Esteem
About 2.9 % (28.76/10 = 2.9) of the variance in “Depression 1988 Recoded” can be accounted for by all four predictors (X1, X2, X3 and X4). The fourth model, which includes all four predictors, resulted in a high F ratio of 23.12244031 with a p <.001.
Model 5: Negative Affect
About 4.1% (40.9/10 = 4.1) of the variance in “Depression 1988 Recoded” can be accounted for by all five predictors (X1, X2, X3, X4 and X5). The fifth model, which includes all five predictors, resulted in a high F ratio of 30.7 with p <.001.
This study has demonstrated an internally valid, comprehensive and interactive model regarding the therapeutic effects of exercise and sporting activities on depression and self-esteem. That it is arguably a cross-section analysis of one culture at a given moment in time should encourage other researchers to extend the program as a longitudinal study or to replicate the method across other cultures and national milieus.
Given that these findings are consistent with prior work on clinical cases of frank.
Replace Table 21 In Appendix B.