Construct Validity of the Revised Anxiety Rating Scale (ARS-2).

Cox, Richard H. ; Robb, Marshall ; Russell, William D. 等

Development of the revised Anxiety Rating Scale(ARS-2), along with reporting concurrent validity coefficients were detailed in Cox and Robb (1998) The purpose of the current research was to demonstrate construct validity of the ARS-2. The ARS-2 was administered to 248 undergraduate intramural basketball players approximately IS mm before the start of a play-off game. Data were analyzed using multiple regression (MR) and multivariate analysis of variance (MANOVA) procedures. In MR, the dependent variable (performance) was conceptualized as the participant's game score divided by the opponent game score. Regressing performance on cognitive state anxiety, somatic state anxiety, and self-confidence) yielded a significant R2 of .04 for men, F(3,184) = 2.72, p = .046. The standardized betas for cognitive anxiety, somatic anxiety, and self-confidence respectively were -.09, .00, and. 17. For the MANOVA, dependent variables were average team scores on cognitive anxiety, somatic anxiety, and self-confidence as measur ed by the ARS-2. The independent variable was game outcome (win/loss). Multivariate analysis of variance procedures (MANOVA) revealed a significant relationship between game outcome and scores on the ARS-2, F(3,46) = 2.84, p = .048. Relative importance of ARS-2 components, in determining game outcome, was estimated through discriminant analysis and the calculation of structure coefficients. Structure coefficients for cognitive anxiety, somatic anxiety, and confidence were -.58, -.07, and .93 respectively, indicating that self-confidence followed by cognitive anxiety were meaningful in predicting game outcome. These results demonstrate the construct validity of the ARS-2, in that winning and losing basketball teams can be predicted as a function of scores on the ARS-2.

The Anxiety Rating Scale (ARS) was developed by Cox, Russell and Robb (1998, 1999) as a short rating scale version of the Competitive State Anxiety Inventory - 2 (CSAI-2) (Martens, Burton, Vealey, Bump, & Smith, 1990). The ARS was not merely conceptualized as a short version of the CSAI-2 with fewer items devoted to each anxiety construct, but as a single statement "Borg Like" (1973) rating scale that would allow an athlete to quickly rate how they felt in terms of precompetitive cognitive state anxiety, somatic state anxiety, and self-confidence. The concurrent validity of the ARS was established with team sport intramural athletes (Cox et al., 1998, 1999) and with individual sport intramural athletes (Cox, Reed, & Robb, 1997). Results of these investigations have shown the ARS to be moderately correlated with the subcomponents of Martens = CSAI-2 (.60 to .70).

The ARS also has been compared favorably with Krane's (1994) modification of the Mental Readiness Form (MRF-L). The MRF-L is a short version of the CSAI-2 originally developed by Murphy, Greenspan, Jowdy, and Tammen (1989). Comparative correlations between the anxiety components of the ARS and MRF-L with the cognitive and somatic state anxiety subscales of the CSAI-2 have consistently favored the ARS (Cox et al., 1997; Cox et al., 1999). As originally developed (Cox et al., 1998), the ARS measured competitive somatic state anxiety and competitive cognitive state anxiety, but not state self-confidence. The measurement of state self-confidence was later added to the ARS using stepwise multiple regression procedures to identify items to be included (Cox et al., 1999).

Most recently, the original ARS was revised and named the Anxiety Rating Scale - 2 (ARS-2). The ARS-2 exhibits stronger concurrent validity than the original scale and eliminates ambiguous wording in response statements (Cox & Robb, 1998). In developing the ARS-2, the original ARS was compared with two potential revised versions relative to concurrent validity with the CSAI-2. Reported correlations for the ARS-2 with the CSAI-2 for cognitive state anxiety, somatic state anxiety, and self-confidence were .67, .69, and .75, respectively. The ARS-2 is composed of three aggregate statements set to a 7-point Likert scale (1=not at all, 2=a little bit, 3=somewhat, 4=moderately so, 5=quite a bit, 6=very much so, 7=intensely so) which allow an athlete to rate how they feel immediately prior to competition. Respectively, the three statements for cognitive state anxiety, somatic state anxiety, and self-confidence read as follows:

1. I feel concerned about performing poorly, choking under pressure, and that others will be disappointed with my performance.

2. I feel jittery, my body feels tense, and my heart is racing.

3. I feel comfortable, secure, and confident about performing well.

In retrospect, it seems clear that the concurrent validity of the original ARS and the revised ARS-2 has been well established (Cox & Robb, 1998; Cox et al., 1997; Cox et al., 1998; Cox et al., 1999). It has been demonstrated that the ARS and ARS-2 are effective in measuring precompetitive anxiety and confidence when compared concurrently with the parent CSAI-2. A next step evolution in the continued testing of the ARS-2 would be to establish the construct validity of the instrument. A test that possesses construct validity should be able to discriminate between two groups of athletes who differ on a hypothetical construct measured by the instrument (Thomas & Nelson, 1996). Theoretically, successful and less successful athletes should differ on precompetitive anxiety and self-confidence (Fenz & Jones, 1972; Highlen & Bennett, 1979; Mahoney & Avener, 1977; Meyers, Cooke, Cullen & Liles, 1979; Wiggins & Brustad, 1996). Superior athletes should exhibit higher levels of precompetitive self-confidence and lower l evels of state anxiety than less accomplished athletes.

Therefore, it is the purpose of this investigation to establish the construct validity of the ARS-2 by testing the ability of the instrument to predict performance and to discriminate among groups of athletes who win or lose an athletic contest. Specifically, it is predicted that (a) team outcome score ratios can be predicted from individual athlete's scores on the ARS-2, and (b) average team scores on the constructs measured by the ARS-2 will differ for winning and losing teams.

Method

Participants

Participants for this research were 188 male and 60 female undergraduates competing in 25 intramural basketball games at a large Midwest University. Use of human subjects for the purpose of research was approved by the appropriate University human subject's committee and anonymity assured.

Instrument

The instrument used for measuring precompetitive state anxiety and self-confidence was the revised Anxiety Rating Scale (ARS-2). The reported concurrent validity of the ARS-2 is .67 for cognitive state anxiety, .69 for somatic state anxiety, and .75 for self-confidence (Cox & Robb, 1998). The ARS-2 was previously described in some detail in the introduction of this article.

Procedures

Starting members of intramural basketball teams were approached approximately 15 mm prior to a play-off competition and asked if they would participate in the study. Because the ARS-2 only takes a few seconds to administer, it was rare for an athlete to decline to participate. In a single contest, starting members of both teams were invited to participate. Because winners advanced to the next round, some players completed the ARS-2 more than once. The ARS-2 was administered 188 times to males and 60 times to females (N248).

Design and Analysis

Data were analyzed using multiple regression (MR) and multivariate analysis of variance (MANOVA) procedures. In MR, the dependent variable (performance) was conceptualized as the participant's game score divided by the opponent's score. Thus, each member of a specific team would have the same performance score for that contest. A performance score of greater than unity would indicate a win, while a score of less than unity would indicate a loss. Very large or very small scores would indicate a large margin of victory or loss. ARS-2 scores on cognitive state anxiety, somatic state anxiety, and self-confidence served as continuous property independent variables. In the MANOVA, dependent variables were average team scores on cognitive state anxiety, somatic state anxiety, and self-confidence as measured by the ARS-2. The independent variable was game outcome (win/loss). In all analyses an alpha level of .05 was adopted.

Results

Multiple Regression Analyses

Means and standard deviations associated with the dependent variable and three independent variables are displayed in Table 1. Simple correlations among independent variables and with the dependent variable of performance are displayed in Table 2. Relatively small correlations were observed between performance and the three subscales of the ARS-2, with the largest being associated with self-confidence. Near collinearity was inspected relative to the three subscales of the ARS-2. As can be observed in Table 2, the correlations among independent variables range from .274 to .530 (ignoring the sign). These correlation coefficients would not be suggestive of problems associated with near collinearity. To confirm the absence of near collinearity among independent variables, variance inflation factors (VIF), tolerance, and condition indexes (CI) were calculated for each variable for each regression analysis. None of the calculated tolerance values approached zero and all of the VIF's were less than 2.00. In additi on, collinearity diagnostics, associated with calculated CI's, confirmed the absence of near collinearity among independent variables.

Separate regression analyses were calculated for male participants, female participants, and male and female participants combined. In all three analyses, performance was regressed upon cognitive state anxiety (cog), somatic state anxiety (som), and self-confidence (conf). The regression analysis for males yielded a significant [R.sup.2] of .0424, F (3,184)=2.72, p =.046. The regression analysis for females yielded a nonsignificant [R.sup.2] of .0273, F(3,56) = .523, p = .6681. Finally, the regression analysis for the combined sample yielded a nonsignificant R2 of .0297, F(3,244),2.49, p=.0612.

Parameter estimates, results of significance tests, and unique variance (squared semi partial correlations) are displayed in Table 3 relative to the significant regression model for males. As can be observed in this Table, the regression coefficients for the two measures of state anxiety were not significant. The regression coefficient for self-confidence was significant, and the squared semi-partial correlation with performance was .0242. These results suggest that of the three subscales of the ARS-2, only self-confidence has a significant effect upon performance.

Multivariate Analysis of Variance

Separate one factor MANOVA's were calculated for male participants, female participants, and a combined sample of male and female participants. In all three MANOVA's, the independent variable was two levels of game outcome (win/loss), and the multivariate dependent variables were cognitive state anxiety, somatic state anxiety, and self-confidence as measured by the ARS-2. The separate MANOVA for males yielded a nonsignifcant Wilks' Lambda statistic, F (3,34) = 1.602, p = .2069, as did the MANOVA for females, F(3,8) = 1.289, p = .3426. The combined sample MANOVA yielded a significant relationship between game outcome and the three dependent variables, F(3,46) 2.843, p = .0479. The Wilks = Lambda was .843 ([Eta.sup.2]= .157). To determine the relative contribution of the dependent variables in the significant relationship, the MANOVA was followed by a descriptive discriminant analysis.

The discriminant function, that results from the descriptive discriminant analysis, may be used to calculate a discriminant score for each participant. The discriminant scores are then correlated with each participant's original dependent variables. These resulting correlations are referred to as structure coefficients and may be used to determine the relative importance of each dependent variable. Structural coefficients greater than .30 are considered to be meaningful (Pedhzur, 1997). In the present case these structure coefficients were based upon the total sums of squares and cross-products matrix and are referred to as total structural coefficients. The total structure coefficients for cognitive state anxiety, somatic state anxiety, and self-confidence were -.5845, -.0635, and .9316 respectively. These coefficients suggest that confidence followed by cognitive state anxiety are meaningful in discriminating between winning and losing basketball teams.

Discussion

The prediction that team outcome score ratios can be predicted from individual athlete's scores on the ARS-2 was supported for males, but not for females or for the combined male/ female sample. An inspection of simple correlations for women between team outcome score ratios and the three subscales of the ARS-2 reveal low and insignificant relationships. The inclusion of the female participants into the combined male and female sample resulted in an insignificant R2 for this model as well. The all male multiple regression model was able to reliably predict team outcome score ratios, but only self-confidence exhibited a significant regression coefficient. With some limitations, the ARS-2 was effective in predicting team outcome score ratios.

The prediction that average team scores on the constructs measured by the ARS-2 would differ for winning and losing teams was supported for the total sample of men and women, but not separately. From a total sample perspective, it was possible to discriminate between winning and losing teams as a function of precompetitive scores on the ARS-2. Based upon a follow-up discriminant analysis and calculation of structure coefficients, it was determined that self-confidence followed by cognitive state anxiety were most important in discriminating between winning and losing teams. Somatic state anxiety contributed very little to the

Ideally a stronger and more consistent relationship between precompetitive scores on the ARS-2 and team performance was desired. Never-the-less this research provided support for the construct validity of the revised Anxiety Rating Scale (ARS-2) in a ecologically valid research setting. The self-confidence subscale and, to some degree, the cognitive state anxiety subscale were generally effective in discriminating between teams that won or lost their basketball contests. At least one attempt to demonstrate the construct validity of the parent CSAI-2 was also less than ideal. Martens, Burton, Vealey, Bump, and Smith (1990) administered the CSAI-2 to 49 male Junior National Golf Tournament participants and correlated subscales with performance scores for nine rounds of tournament golf. No significant correlations were observed between golf performance and the CSAI-2 subscales of cognitive state anxiety, somatic state anxiety, or self-confidence. Supporting evidence for the construct validity of the CSAI-2 was forthcoming in subsequent investigations (Martens et al., 1990).

In the present investigation, the construct validity of the revised Anxiety Rating Scale (ARS-2) was established using intramural athletes participating in the team sport of basketball. As a next step in establishing the construct validity of the ARS-2, it is recommended that it be tested using individual sport athletes as well as different age groups. While not without its limitations, the ARS-2 is a viable short form measuring precompetitive state anxiety and self-confidence when minimizing time involvement and athlete distraction are of primary concern.

References

Borg, G.A.V .(1973). Perceived exertion: A note on 'history' and methods. Medicine and Science in Sports, 5, 90-93.

Cox, R.H., Reed, C., & Robb, M. (1997). Comparative validity of the MRF-L and ARS competitive state anxiety rating scales for intramural athletes competing in six individual sports. Research Quarterly for Exercise and Sport. 68(1), Suppl. A-101.

Cox, R.H., & Robb, M.( 1998). Concurrent validity of the revised anxiety rating scale. Journal of Applied Sport Psychology, 10 (supplement), S 160.

Cox, R.H., Russell, W.D., & Robb, M. (1998). Development of a CSAI-2 Short form for assessing competitive state anxiety during and immediately prior to competition. Journal of Sport Behavior, 21,31-40.

Cox, R.H., Russell, W.D., & Robb, M. (1999). Comparative concurrent validity of the MRF-L and ARS competitive state anxiety rating scales for volleyball and basketball. Journal of Sport Behavior, 22, 310-320.

Fenz, W.D., & Jones, G.B. (1972). Individual differences in physiologic arousal and performance in sport parachutists. Psychosomatic Medicine, 34, 1-8.

Highlen, P.S., & Bennett, B.B. (1979). Psychological characteristics of successful and unsuccessful elite wrestlers: An exploratory study. Journal of Sport Psychology, 1, 123-127.

Krane, V. (1994). The mental readiness form as a measure of competitive state anxiety. The Sport Psychologist, 8, 189-202.

Mahoney, M.J., & Avener, M. (1977). Psychology of the elite athlete: An exploratory study. Cognitive Therapy and Research, 3, 361-366.

Martens, R., Burton, D., Vealey, R.S., Bump, L.A., & Smith, D.E.(1990). Development and validation of the competitive state anxiety inventory-2. In R. Martens, R.S. Vealey, & D. Burton (Ed's.), Competitive Anxiety in Sport (pp. 117-190). Champaign, IL: Human Kinetics Books.

Meyers, A.W., Cooke, C.J., Cullen, J., & Liles, L. (1979). Psychological aspects of athletic competitors: A replication across sports. Cognitive Therapy and Research, 3, 361-366.

Murphy, S.M., Greenspan, M., Jowdy, D., & Tammen, V. (1989). Development of a brief rating instrument of competitive anxiety: Comparisons with the Competitive State Anxiety Inventory-2. Proceedings of the Association for the Advancement of Applied Sport Psychology (p.82). Seattle, WA.

Thomas, J.R., & Nelson, J.K. (1996). Research methods in physical activity (3rd edition). Champaign, IL: Human Kinetics.

Wiggins, M.S., & Brustad, R.J. (1996). Perception of anxiety and expectations of performance. Perceptual and Motor Skills, 83, 1071-1074.

 Means and Standard Deviations Associated
 with Variables in Multiple Regression Model
 Variables
 Cognitive Somatic
 Performance Anxiety Anxiety Confidence
Gender N Mean SD Mean SD Mean SD Mean
Male 188 1.06 .436 2.76 1.57 2.62 1.41 5.15
Female 60 1.09 .468 2.48 1.67 2.00 1.21 4.93
Combined 248 1.07 .443 2.69 1.59 2.47 1.39 5.10
Gender SD
Male 1.26
Female 1.65
Combined 1.36
 Correlations among ARS-2 Subscales and with Performance
 as Associated with Multiple Regression Analyses
 Variables
 Cognitive Somatic Self-
Gender Variable Performance Anxiety Anxiety Confidence
 Performance 1.000
 Cognitive Anxiety -.132 1.000
Male Somatic Anxiety -.086 .463 [**] 1.000
 Self-Confidence -.188 [**] -.274 [**] -.285 [**] 1.000
 Performance 1.000
 Cognitive Anxiety -.123 1.000
Female Somatic Anxiety .038 .437 [**] 1.000
 Self-Confidence .068 -.530 [**] -.512 [**] 1.000
 Performance 1.000
 Cognitive Anxiety -.l31 [*] 1.000
Combined Somatic Anxiety -.062 .460 [**] 1.000
 Self-Confidence .148 [*] -.343 [**] -.318 [**] 1.000
(*.)Significant at .05
(**.)Significant at .01
 Parameter Estimates Associated with
 Multiple Regression Model for Male Participants
 Parameters [*]
 Stand. Unique
Variable beta Error T-Ratio Prob Beta [**] Variance
Cognitive Anxiety -.0242 .0229 -1.057 .2920 -.0872 .0058
Somatic Anxiety .0003 .0255 0.012 .9907 .0010 .0000
Self-Confidence .0570 .0264 2.156 .0324 .1646 .0242
(*.)Intercept of unstandardized model equals .8370
(**.)Standardized regression coefficient