spanish adaptation and analysis by structural equation modeling...

SpaniSh adaptation and analySiS by Structural Equation ModEling of an inStruMEnt for

Monitoring ovErtraining: thE rEcovEry-StrESS quEStionnairE (rEStq-Sport)

René González-Boto, Alfonso sAlGueRo, ConCepCión tueRo, And sARA MáRquez

University of León, SpainMiChAel KellMAnn

University of Queensland, Australia

This study was aimed at examining the psychometric properties of a Spanish version of the Recovery-Stress Questionnaire for Athletes (RESTQ-Sport; Kellmann & Kallus, 2000), and determining relationships between specific and overall dimensions of stress and recovery. A sample of Spanish athletes from various sports completed the instrument. Two factors were obtained in both the general (Sport Non-Specific Stress and Sport Non-Specific Recovery) and the specific module of the questionnaire (Sport Specific Stress and Sport Specific Recovery). A recursive model with a satisfactory fit, in which all factors would explain a total stress dimension, a total recovery dimension and a stress-recovery state, was developed. Results support the internal validity and reliability of the Spanish version of the RESTQ-Sport, and allow for the establishment of relationships among the various different constructs that form the basis of the questionnaire.

Keywords: overtraining, RESTQ-Sport, stress, recovery, validity, reliability, Spanish.

Performing regular physical activity in a variety of settings is said to provide favorable consequences for a large variety of health outcomes (Miké, 2003).

SOCIAL BEHAVIOR AND PERSONALITY, 2008, 36(5), 635-650© Society for Personality Research (Inc.)

635

René González-Boto, Alfonso Salguero, Concepción Tuero, and Sara Márquez, Faculty of Physical Activity and Sport Sciences, University of León, Spain; and Michael Kellmann, School of Human Movement Studies and Psychology, University of Queensland, Australia.This study was supported by research grants from the Acción Estratégica sobre el Deporte, and the Regional Government of Castilla y Leon, Spain.Appreciation is due to reviewers including: Jarek Mäestu, Department of Psychology, Institute of Sport Pedagogy and Coaching Sciences, University of Tartu, Jakobi 5, Tartu, Estonia 50090, Email: [email protected] address correspondence and reprint requests to: Sara Márquez, Facultad de Ciencias de la Actividad Física y el Deporte, Universidad de León, Campus Universitario, 24071 León, Spain. Phone: +34 629 543389; Fax: +34 987 291267; Email: [email protected]

RECOVERY-STRESS QUESTIONNAIRE636

But occasionally some individuals take physical fitness or the pursuit of optimal performance too far. They are either committed exercisers who overtrain to the point of burnout, or sedentary individuals who begin to embrace a workout program too aggressively. Overtraining is a consequence of exercise beyond the body’s ability, when training intensity, duration, or volume surpasses the recovery time being offered to the body (Kellmann, 2002). This can be hard for many people to grasp, as they believe that if some training is good, more must be better. In addition to a decline in physical performance, signs and symptoms such as dwindling enthusiasm for working out, increases in resting heart rate and resting blood pressure, continuous muscle or joint soreness, decrease in appetite and body weight, disturbed sleep, and increased irritability, anxiety, or depression, can appear (Urhausen & Kindermann, 2002). Thus, overtraining can be harmful and lead to serious physical and psychological problems, requiring special care and a high degree of attention from health care professionals.

Most definitions link overtraining with an imbalance between factors favoring stress and those aiding recovery (Kellmann, 2002), so that individuals suffer premature fatigue when faced with exercise, do not achieve complete recovery after efforts, and show limited or even reduced performance. This imbalance has generally been identified with situations involving exercise that relate to the dynamics of learning to manage workloads during the process of adaptation (Lehmann, Foster, Gastmann, Keizer, & Steinacker, 1999). Perception of effort, mood states and several behavioral indicators have been proven to be affected by overtraining and hence are the psychological constructs that have most often been taken into account in the evaluation of overtraining. Garcin, Fleury, and Billat (2002), using the Perceived Exertion Scale (RPE; Borg, 1970), found that overtrained subjects had a perception of greater effort than individuals who were not overtrained. However, the effectiveness of using scales that measure effort in isolation for the diagnosis of overtraining has been questioned (Urhausen & Kindermann, 2002). By means of the Profile of Mood States (POMS; McNair, Lorr, & Droppleman, 1971), different researchers have been able to demonstrate that overtrained subjects experienced a rise in depression levels, anger, and fatigue, combined with decreased vigor (Berglund & Säfström, 1994). However, other studies have not confirmed the suitability of this tool for following up the level of stress and overtraining (Martin, Andersen, & Gates, 2000).

Kellmann and Kallus (2000) developed the Recovery-Stress Questionnaire for Athletes (RESTQ-Sport), which measures both current perceived stress and recovery in a multidimensional way, offering an approximate idea both of the stressing agents and of the recovery strategies that are being used. Therefore, it may be more effective than the previously used Borg ratio scale and POMS, which both focus mainly on the stress component. The RESTQ-Sport is based on the theoretical model proposed by Kellmann (2002), which describes the

RECOVERY-STRESS QUESTIONNAIRE 637

interrelations of stress states and recovery demands. The basic assumption is that as stress increases, increased time of recovery is necessary. The situation of balance or imbalance is determined by the individual’s capacity to make use of resources necessary for recovery, and by their strategies for coping with situations of stress. If the resources are appropriate, the subject may react satisfactorily and cope well with stress without needing to use additional recovery activities. However, in the absence of recovery, or in the situation of under-recovery, the athlete initiates a process of stress which can lead to levels beyond the threshold of the optimal margin. The model goes on to consider the analysis of overtraining on the basis of the recovery-stress state, considered as a multidimensional concept which represents the magnitude of physical and psychological stress in an individual, as well as individual capacity to employ recovery strategies and which strategies are used.

Various studies have highlighted the validity and reliability of the RESTQ-Sport questionnaire (Kellmann & Kallus, 2000, 2001). Further research with the Portuguese and Estonian versions of the RESTQ-Sport supports the psychometric properties of the survey as well (Costa & Samulski, 2005; Jürimäe, Mäestu, Purge, Jürimäe, & Soot, 2002; Mäestu, Jürimäe, Kreegipuu, & Jürimäe, 2006). The RESTQ-Sport has been used to evaluate the effects of rapidly increased training volume on stress-recovery state and its relationship with performance and biochemical markers (Jürimäe, Mäestu, Purge, & Jürimäe, 2004). Recent research has demonstrated that the RESTQ-Sport may be a practical tool for recognizing overtraining in its early stages (Coutts, Wallace, & Slattery, 2007).

The general purpose of the work being reported here was to develop a Spanish version of the RESTQ-Sport. In terms of specific objectives, it was intended to use the data collected to analyze internal validity and reliability, as well as the relationship between the various components of the theoretical model on which the questionnaire is based. This implies an improvement over previous versions of the questionnaire in other languages, in which such relationships have not been empirically established.

MEthod

ParticiPants

The sample was composed of 294 subjects (53% male and 47% female), ranging in age from 18 to 24 years (M = 21.0, SD = 2.0). Participants in both individual (47%) and team sports (53%) were included, representing individual sports such as athletics (6.2%), swimming (6.2%), cycling (4.6%) or judo (3.1%) and team sports such as basketball (19.6%), soccer (16.0%), rugby (7.2%) and indoor soccer (6.2%).


Measures

The Recovery-Stress Questionnaire for Athletes (RESTQ-Sport; Kellmann & Kallus, 2000, 2001) consists of 76 items. A Likert-type scale, with values ranging from 0 (never) to 6 (always), indicates how often the respondent participated in various activities during the past three days and nights. The measure includes twelve scales from the general Recovery-Stress Questionnaire, developed by Kallus (1995), which assess various stressing agents of a general nature and general recovery activities during day-to-day life. To focus more on stress and recovery in sports, seven additional sports-specific scales were developed. These investigate aspects complementary to stress that are derived from the area of sport and assess specific recovery activities derived from the sport context. The Spanish version of the RESTQ-Sport was translated through a parallel back-translation procedure by individuals specialized in sports psychology (Van de Vijver & Hambleton, 1996). In addition to answering the RESTQ-Sport questionnaire, participants also completed a sociodemographic questionnaire.

Procedure

The authors contacted the directors of sports clubs, explained the nature of the study, and asked permission to interview individuals from membership lists. Of the original sample of people contacted, 73% agreed to participate in the study. They were informed that participation was voluntary and informed consent was obtained. They were told what the exact purpose of the study was and a strong emphasis was put on data confidentiality. Participants were asked to answer each item as honestly as possible and to complete their questionnaires individually.

data analysis

A principal component analysis was performed using the responses from the 76 items of RESTQ-Sport. Only those factors with eigenvalues of 1.0 or higher were retained for the final rotation. The stability and applicability of the proposed underlying factor structure was investigated by carrying out a confirmatory factor analysis using AMOS Version 5.0 (Arbuckle, 2003). Before testing the hypothesized underlying structure of the RESTQ-Sport, univariate skewness and kurtosis of items were examined. The multivariate normal distribution of responses was also examined utilizing Mardia’s coefficient of multivariate kurtosis (West, Finch, & Curran, 1995).

The Maximum Likelihood (ML) method was used to produce the covariance matrix in estimating the relationships between variables. Evaluation of model fitting solely by using the chi-square statistics, although representing the main measure of a global model fit, increases the likelihood of Type-II error. Thus, the recommendations of Hu and Bentler (1999) and McDonald and Ringo Ho (2002) were followed, and this was complemented by relative fit indices, such


as the goodness of fit index (GFI), the comparative fit index (CFI), the Tucker-Lewis index (TLI), the normed fit index (NFI), the standardized root mean square residual (SRMR) and the root mean square of approximation (RMSEA). Because chi-square is related to the degrees of freedom in the model, in accordance with the proposal by Wheaton, Muthén, Alwin, and Summers (1977), results are expressed as the minimum value of the discrepancy divided by its degrees of freedom (CMIN/df). The reliabilities of factors and scales were assessed using Cronbach’s index (Cronbach, 1951) and interscale correlations.

rESultS

distribution of the restQ-sPort iteMs

The univariate skewness values of the items ranged from -1.22 to 2.48 and their univariate kurtosis ranged from -1.48 to 5.47. None of the items were eliminated, as no outliers occurred in relation to the intervals proposed by Chou and Bentler (1995) and by West, Finch, and Curran (1995). Multivariate normality of the data was assessed using Mardia’s coefficient of multivariate kurtosis, yielding a figure of 248.97 with a critical ratio of 22.32. Results revealed that data violated the assumption of a multivariate Gaussian distribution. Thus, it was decided to adjust the p value of the chi-square statistic with the boot strapping procedure of Bollen and Stine (1993). Studies undertaken by Nevitt and Hancock (1997) with non-normal data lent support to this decision by revealing that Bollen-Stine corrected chi-square performed better than the Satorra-Bentler scaled chi-square statistic with small samples.

PrinciPal coMPonent analysis

Two factors were identified in both the general and sport-specific modules (Table 1). In the general module, the two factors accounted for 41% of the variance. The first corresponded to Sport Non-Specific Stress (SNSS), in which all stress items had a loading. Except for one item (load 0.37), all items had loadings in the range from 0.44 to 0.75. The second factor, Sport Non-Specific Recovery (SNSR), included 20 items with scores ranging from 0.41 to 0.80 in 17 out of 20 items. In the sport-specific items module, the two factors accounted for 39% of the variance. Twelve items, with scores ranging from 0.41 to 0.64, had a weighting in the Sport Specific Stress (SSS) factor. The second factor, Sport Specific Recovery (SSR), comprised 16 items, with scores ranging from 0.43 to 0.82.


Table 1General and sPecific factors of the restQ-sPort

NON-SPeCIFIC SPORT FaCTORS Stress Factor (SNSS) Recovery Factor (SNSR)Item Load Item Load2. I did not get enough sleep 0.47 3. I finished important tasks 0.414. I was unable to concentrate well 0.59 6. I laughed 0.555. Everything bothered me 0.61 9. I felt physically relaxed 0.427. I felt physically bad 0.53 10. I was in good spirits 0.768. I was in a bad mood 0.57 13. I felt at ease 0.7211. I had difficulties in concentrating 0.58 14. I had a good time with my friends 0.6412. I worried about unresolved problems 0.56 17. I was successful in what I did 0.6915. I had a headache 0.57 19. I fell asleep satisfied and relaxed 0.5016. I was tired from work 0.63 23. I visited some close friends 0.2918. I couldn’t switch my mind off 0.62 27. I had a satisfying sleep 0.4520. I felt uncomfortable 0.65 29. I felt physically fit 0.5921. I was annoyed by others 0.62 33. I had fun 0.6922. I felt down 0.70 34. I was in a good mood 0.7924. I felt depressed 0.72 36. I slept restlessly 0.3825. I was dead tired after work 0.70 38. I felt as if I could get everything done 0.6226. Other people got on my nerves 0.75 41. I made important decisions 0.4128. I felt anxious or inhibited 0.61 43. I felt happy 0.7930. I was fed up with everything 0.72 46. My sleep was interrupted easily 0.2031. I was lethargic 0.59 47. I felt content 0.8032. I felt I had to perform well 49. I had some good ideas 0.52 in front of others 0.3735. I was overtired 0.67 37. I was annoyed 0.67 39. I was upset 0.72 40. I put off making decisions 0.44 42. I felt physically exhausted 0.62 44. I felt under pressure 0.57 45. Everything was too much for me 0.72 48. I was angry with someone 0.53 Eigenvalue 14.53 Eigenvalue 5.16 % of Variance 30.27 % of Variance 10.75Mean weight 0.61 Mean weight 0.56 Cronbach α 0.94 Cronbach α 0.90

SPeCIFIC SPORT FaCTORSStress Factor (SSS) Recovery Factor (SSR) Item Load Item Load50. Parts of my body were aching 0.57 52. I was convinced I could achieve 51. I could not get rest during breaks 0.60 my set goals during performance 0.7153. I recovered well physically 0.69 55. I accomplished many worthwhile 54. I felt burned out by my sport 0.62 things in my sport 0.6057. My muscles felt stiff or tense 56. I prepared myself mentally for during performance 0.41 performance 0.7258. I had the impression there were 59. I was convinced that I could achieve too few breaks 0.64 my performance at any time 0.64


Table 1 continued

Item Load Item Load63. I felt emotionally drained from 60. I dealt very effectively with my performance 0.43 team-mates’ problems 0.5364. I had muscle pain after performance 0.51 61. I was in a good condition physically 0.7066. Too much was demanded of me 62. I pushed myself during performance 0.69 during the breaks 0.53 65. I was convinced that I performed well 0.6568. I felt that I wanted to quit my sport 0.60 67. I psyched myself up before72. The breaks were not at the right performance 0.74 times 0.43 69. I felt very energetic 0.8273. I felt vulnerable to injuries 0.41 70. I easily understood how my76. I felt frustrated by my sport 0.63 team-mates felt about things 0.43 71. I was convinced that I had trained well 0.63 74. I set definite goals for myself during performance 0.57 75. My body felt strong 0.79 77. I deal with emotional problems in my sport very calmly 0.47Eigenvalue 3.71 Eigenvalue 7.31% of Variance 13.25 % of Variance 26.09Mean weight 0.53 Mean weight 0.65Cronbach α 0.77 Cronbach α 0.91

structural Models

An initial analysis tested a 76-item four-factor model. Endogenous items’ residuals were allowed to correlate freely but it was decided not to allow free correlations between the residuals of exogenous variables. After the specification and identification stages an over identified model emerged. After the estimation stage the factor rating of item 23 did not reach significance (p > 0.05). Factor ratings were significant (p < 0.05) for 9 items and highly significant (p < 0.001) for the remaining items. Correlations ranged from 0.20 to 0.88. All indices were unsatisfactory, revealing a weak fit to the data: CMIN/df = 2.12 (p < 0.05), GFI = 0.53, TLI = 0.62, CFI = 0.63, NFI = 0.48, SRMR = 0.10 and RMSEA = 0.08. Modification Indices (MI) were examined to detect where the model could be improved. It was decided to delete the items with the highest index values in order to reach a fit that was as satisfactory as possible. Twenty-one items were selected. This standardized solution reflected highly significant regression ratings for all items (p < 0.001) with values ranging from 0.26 to 0.87. The value for CMIN/df was 1.23, the BS chi-square was not significant (p = 0.413) and all the other indices (GFI = 0.90, TLI = 0.97, CFI = 0.97, NFI = 0.88, SRMR = 0.06, and RMSEA = 0.03) reflected a more satisfactory fit. Using the previous model as a basis, a new hypothesis was proposed, with two hypothetical latent constructs - a total stress dimension (TS) and a total recovery (TR) dimension,


which describe the overall state of stress and recovery (S-R State), and how they would affect the four first-order factors. For this purpose two new endogenous variables were introduced at a second level to explain the relationship with the four factors established, together with an exogenous latent variable on a third level to establish a relationship with the two new dimensions. A recursive model was developed with fit indices (CMIN/df = 1.23 with pBS chi-square = 0.413; GFI = 0.90, TLI = 0.97, CFI = 0.97, NFI = 0.91, SRMR = 0.06, and RMSEA = 0.03) that reflected a satisfactory fit. In the standardized solution scores for regression between the four factors and total stress and recovery dimensions were 0.50

23

Figure 1

Standardized Parameter Estimates of the RESTQ-Sport Based on Model 3. SNSS: Sport Non-

Specific Stress, SNSR: Sport Non-Specific Recovery, SSS: Sport Specific Stress, SSR: Sport

Specific Recovery, TS: Total Stress; TR: Total Recovery; S-R State: Level of Stress and

Recovery

SNSS

.55rs_22 e22

.74 .72rs_24 e24

.85.41

rs_28 e28.64.54rs_26 e26

.73

.55rs_39 e39

.74

.21rs_40 e40

.46

SNSR

,16rs_49 e49

.37rs_6 e6

.56rs_33 e33

.54rs_13 e13

.75rs_34 e34

.76rs_47 e47

.39.61.75.73.86.87

.11rs_36 e36

.33

SSS

.61rs_54 e54

.78.65

rs_68 e68.81.66

rs_76 e76.82

.07rs_57 e57

.26

SSR

.67rs_69 e69

.22rs_77 e77

.42rs_52 e52

.41rs_67 e67

.82.47.65.64

-.34

.76

-.24 -.41

.56

-.52

.81

TS

.80

TR

.50

.52

.69

.25

eTS

eTR

S-RState

.82

.79

SNSS

.55rs_22 e22

.74 .72rs_24 e24

.85.41

rs_28 e28.64.54rs_26 e26

.73

.55rs_39 e39

.74

.21rs_40 e40

.46

SNSR

,16rs_49 e49

.37rs_6 e6

.56rs_33 e33

.54rs_13 e13

.75rs_34 e34

.76rs_47 e47

.39.61.75.73.86.87

.11rs_36 e36

.33

SSS

.61rs_54 e54

.78.65

rs_68 e68.81.66

rs_76 e76.82

.07rs_57 e57

.26

SSR

.67rs_69 e69

.22rs_77 e77

.42rs_52 e52

.41rs_67 e67

.82.47.65.64

-.34

.76

-.24 -.41

.56

-.52

.81

TS

.80

TR

.50

.52

.69

.25

eTS

eTR

S-RState

.82

.79

Figure 1: Standardized Parameter Estimates of the RESTQ-Sport Based on Model 3. SNSS: Sport Non-Specific Stress, SNSR: Sport Non-Specific Recovery, SSS: Sport Specific Stress, SSR: Sport Specific Recovery, TS: Total Stress; TR: Total Recovery; S-R State: Level of Stress and Recovery.


(TS-SNSS), 0.52 (TS-SSS), 0.69 (TR-SSR) and 0.25 (TR-SNSR). Between the total stress and total recovery dimensions and the level of stress and recovery, the figures were 0.82 and 0.79 respectively. The squared multiple correlations for TS and TR were 0.81 and 0.80, respectively (Figure 1).

reliability and interfactor correlations

On the basis of the results from the principal component factor analysis, Cronbach’s index of internal consistency was calculated for the four factors in the questionnaire. Values ranged from 0.77 for the SSS factor to 0.94 for the SNSS factor. Cronbach alpha was in all cases higher than the recommended minimum 0.70 for psychological scales (Nunnally & Bernstein, 1994). Results for all 19 scales of the RESTQ-Sport are shown in Table 2. Most of the scales achieved alpha coefficients equal or higher than 0.70, although some showed moderately low values. Interfactor correlations based on a two-stage sequential confirmatory factor analysis are shown in Table 3. Both stress factors and recovery factors, respectively, showed a moderately high positive correlation among themselves. A moderate negative correlation was noted when stress and recovery factors were correlated with each other.

Table 2subscales of the restQ-sPort, Mean scores (M), standard deviations (SD) and

cronbach alPhas

Scale Stress/Recovery i M SD Cronbach α

1. General Stress s 4 1.28 1.35 0.852. Emotional Stress s 4 1.40 1.10 0.753. Social Stress s 4 1.27 1.19 0.794. Conflicts/Pressure s 4 2.46 1.33 0.675. Fatigue s 4 2.35 1.41 0.776. Lack of Energy s 4 1.81 1.10 0.707. Physical Complaints s 4 1.58 1.10 0.668. Success r 4 2.73 1.12 0.639. Social Recovery r 4 3.69 1.15 0.6310. Physical Recovery r 4 2.95 1.20 0.7111. General Well-Being r 4 3.84 1.39 0.9112. Sleep Quality r 4 3.41 1.60 0.8213. Disturbed Breaks s 4 1.59 1.18 0.6714. Emotional Exhaustion s 4 1.13 1.24 0.7715. Injury s 4 2.53 1.27 0.5416. Being in Shape r 4 3.32 1.51 0.8817. Personal Accomplishment r 4 3.17 1.16 0.6318. Self-Efficacy r 4 3.17 1.34 0.7319. Self-Regulation r 4 3.74 1.39 0.78

Note: s: stress. r: recovery. i: number of items. Values for mean scores, standard deviation and Cronbach alpha.


Table 3correlation between the restQ-sPort factors

Interfactor correlations

SNSS SNSR SSS

SNSR -0.52 - SSS 0.56 -0.42 -SSR -0.24 0.76 -0.34

Note: SNSS: Sport Non-Specific Stress, SNSR: Sport Non-Specific Recovery, SSS: Sport Specific Stress, SSR: Sport Specific Recovery.

diScuSSion

After an analysis of the components extracted, it was observed that there was a distribution of the items into four clearly defined factors associated with a stress component and a recovery component. This factorial structure has been demonstrated in the German and English (Kellmann & Kallus, 2000, 2001) versions of the RESTQ-Sport. The variables loading the factors used in those studies were the scores obtained on the 19 scales, not directly those for the 76 items in the questionnaire. In this study there were factorial loadings with values under 0.4 for some items. Although different authors have established the minimum weight of items per factor as being 0.4 (Thomas & Nelson, 2001), there may be others which are equally acceptable. Thus, in a study similar to this (Waite, Gansneder, & Rotella, 1990), the value was set at 0.3, and Kellmann and Kallus (2000, 2001) fixed the criterion in values greater than 0.25. Taking into account the recommendations of the International Test Commission guidelines (Van de Vijver & Hambleton, 1996) about fairness in tests when they are used in more than one language, we applied a rigorous methodology to the development of the questionnaire and issues of content, culture and language to avoid mistakes with the items. Even so, it is possible that there were difficulties of translation related with culture issues that we were not aware of, which could mean some items were not correctly adapted.

Despite this, the Cronbach alpha coefficients demonstrated acceptable internal consistency for all the factors, as they were higher than the minimum value proposed by Nunnally (1978). The total Cronbach alpha value for all items of the questionnaire in this study (α = 0.89) was somewhat greater than the minimum figure required by Kellmann and Kallus (2001), and all of them were as good as or better than the values achieved for similar tools (Prapavessis, Grove, Maddison, & Zillmann, 2003). On the basis of these data it was decided not to exclude any item from the factorial structure to maintain the initial composition of the questionnaire. The stress and recovery factors are tightly linked to the 19 scales of the questionnaire. In the present research moderately high coefficients


of reliability were obtained in most of the 19 scales, with Cronbach alpha values equal to or higher than 0.7. These results were similar to those arising from the studies by Costa and Samulski (2005) and Kellmann and Kallus (2000, 2001), and slightly lower than the scores reported by Mäestu et al. (2006). The lowest reliability values achieved in some scales may be explained by different reasons. One may be related to the construction of the questionnaire, because the general module of the RESTQ-Sport is based on the Recovery-Stress Questionnaire from Kallus (1995), who pointed out that the meaning of the general formulated dimension could be different for athletes compared to nonathletes. Another explanation may relate to the nature of the sample. It should be noted that the number of different sports specialties was quite large, and both experience in sports and amount of training were relatively heterogeneous. This could have influenced the consistency of certain scales due to the different meaning of specific items for different subjects. A similar explanation was pointed out by Kellmann and Kallus (2000, 2001) and Costa and Samulski (2005), who found sample-dependent lower values for some reliability coefficients during daily practice. An item such as “I easily understood how my teammates felt about things” (α = 0.63), can be of great relevance in subjects who practice team sports, but is likely to be of less relevance to those engaged in individual sports. However, consistency was substantially greater for items that were probably more applied in character and thus more easily identifiable for the subjects. It would be worthwhile to produce future adaptations of the RESTQ-Sport as a function, for example, of the type of exercise and sports practice. Besides what has already been noted, it would appear that the reliability values for scales in this study are satisfactory enough to achieve a multidimensional knowledge of stress and recovery. Some authors assert that the observed values may still be considered acceptable for scales with a limited number of items (Hair, Anderson, Tatham, & Black, 1998), even though we must be cautious in the interpretation of some of our lower scales.

A further objective of this research was to find out whether, in the light of the data observed, there was any relationship among the various different constructs that form the basis of the questionnaire. A structural model was built up from a CFA with four specific factors, which related to the overall dimensions of stress and recovery and the stress-recovery state of the sport practitioners. This analysis implies an improvement from previous versions of the questionnaire in other languages, giving empirical support to the multidimensional model of overtraining. In the initial CFA model a rather limited fit was achieved. Batista and Coenders (2000) state that the size of samples influences the process of estimation, so that the larger the number of variables studied or indicators per factor, the larger the sample studied should be. These researchers propose 200 to 500 subjects per sample in adjustments using Maximum Likelihood, and


our sample is within this range. On the basis of the Modification Indexes (MI), several observable variables were eliminated until a more effective model was obtained. We tried to keep the number of items eliminated from the model as small as possible and to leave at least three per factor, in accordance with the recommendations by Hatcher (1994). Fewer items would have been deleted if residuals had been allowed to correlate based on MI. However, such a procedure would have enhanced the risk of capitalizing on chance and also threatened the external validity of the model. As suggested by Byrne (2001), correlated residuals often signify high content overlap or redundancy between items. The elimination of these items was appropriate as it minimized content redundancy and shortened the questionnaire significantly, which is very convenient in competitive sport settings, without affecting its content broadness and relevance. After the removal of items, the values of fit indices improved on a par with the values proposed as good and/or acceptable in the literature. There could be other competitive models, depending on a number of different items, but this is a “generalizable” model that enables determination of the causal relationships with upper constructs. In any case, the above findings do not necessarily mean that the original model of 76 items and 19 scales should be abandoned and further work with a larger sample size must be undertaken to confirm the necessity of the removal of items.

Moreover, the correlations between factors were positive and fairly high among stress factors and among recovery factors, while they were negative but equally acceptable when stress factors and recovery factors were correlated. These results suggest an acceptable level of independence between the stress and recovery dimensions, and support the unidimensional nature of the specific and nonspecific stress and recovery factors making up the questionnaire. The structure of the questionnaire reflects the relationship between the processes of stress and of recovery described in the literature as well (Coffey, Leveritt, & Gill, 2004). The correlations noted between the specific factors and the overall dimensions of stress and recovery were acceptable, but it is more important to emphasize that through the structural model proposed it was possible to demonstrate that 81% of the variability of TS was explained by the influence of the SNSS and SSS factors. Eighty percent of the variability of TR was accounted for by the influence of the SNSR and SSR factors, while 88% of the variability of the level of stress and recovery was explained by the influence of TS and TR. There was no experimental proof available until now of the existence of a causal relationship between all these constructs, but the data from this study tend to confirm the existence of two dimensions of overall stress and of overall recovery, which are determined multicausally by aspects both linked and not linked to the practice of a sport. These two dimensions in their turn explain the stress-recovery state of the athletes. Kallus and Kellmann (2000) claim that this indicator is directly related to the achievement of objectives during training sessions or competitions. Hence,


we suggest that the RESTQ-Sport is an innovatory contribution in assessing the recovery-stress state of athletes.

From the present study it may be concluded that the Spanish version of the Recovery-Stress Questionnaire for Athletes shows factorial validity and internal consistency. It proved possible to demonstrate that the various elements of stress and recovery that were evaluated, whether specific to sport practice or not, determined the total amount of stress and the overall recovery experienced, and these together explained the current stress-recovery state in our subjects. Although our research represents a first stage in demonstrating construct validity of the measure, and further work is required to confirm external validity and to investigate the extent to which scores on the questionnaire relate behaviors in line with theoretical predictions, results obtained support the utility of the RESTQ-Sport for the assessment of recovery-stress states.

rEfErEncES

Arbuckle, J. L. (2003). AmosTM 5.0: Update to the Amos user’s guide. Chicago: SmallWaters Corporation.

Batista, J. M., & Coenders, G. (2000). Structural equation modelling. Models for the analysis of causal relationships. Madrid: La Muralla & Hespérides.

Berglund, B., & Säfström, H. (1994). Psychological monitoring and modulation of training load of world-class canoeists. Medicine and Science in Sports and Exercise, 26, 1036-1040.

Bollen, K. A., & Stine, R. A. (1993). Bootstrapping goodness-of-fit criteria in structural equation modeling. In K. A. Bollen & J. S. Long (Eds.), Testing structural equation models (pp. 195-226). Newbury Park, CA: Sage.

Borg, G. (1970). Perceived exertion as an indicator of somatic stress. Scandinavian Journal of Rehabilitation Medicine, 2, 92-98.

Byrne, B. M. (2001). Structural equation modeling with AMOS. Basic concepts, applications, and programming. Mahwah, NJ: Lawrence Erlbaum.

Chou, C. P., & Bentler, P. M. (1995). Estimates and tests in structural equation modeling. In R. H. Hoyle (Ed.), Structural equation modeling (pp. 37-55). Thousand Oaks, CA: Sage.

Coffey, V., Leveritt, M., & Gill, N. (2004). Effect of recovery modality on 4-hour repeated treadmill running performance and changes in physiological variables. Journal of Science and Medicine in Sport, 7, 1-10.

Costa, L. O. P., & Samulski, D. M. (2005). Processo de validação do Questionário de Estress e Recuperação para Atletas (RESTQ-Sport) na Lengua Portuguesa [Validation process of the Recovery-Stress Questionnaire for Athletes (RESTQ-Sport) in Portuguese]. Revista Brasileira de Ciencia e Movimento, 13, 79-86.

Coutts, A. J., Wallace, I. K., & Slattery, K. M. (2007). Monitoring changes in performance, physiology, biochemistry, and psychology during overreaching and recovery in triathletes. International Journal of Sports Medicine, 28, 125-134.

Cronbach, L. J. (1951). Coefficient alpha and the internal structure of tests. Psychometrica, 16, 297-333.

Garcin, M., Fleury, A., & Billat, V. (2002). The ratio HLa: RPE as a tool to appreciate overreaching in young high-level middle-distance runners. International Journal of Sports Medicine, 23, 16-21.


Hair, J. F., Anderson, R. E., Tatham, R. L., & Black, W. C. (1998). Multivariate data analysis. Upper Saddle River, NJ: Prentice Hall.

Hatcher, L. (1994). A step-by-step approach to using the SAS system for factor analysis and structural equation modeling. Cary, NC: SAS Institute.

Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1-55.

Jürimäe, J., Mäestu, J., Purge, P., & Jürimäe, T. (2004). Changes in stress and recovery after heavy training in rowers. Journal of Science and Medicine in Sport, 7, 334-339.

Jürimäe, J., Mäestu, J., Purge, P., Jürimäe, T., & Soot, T. (2002). Relations among heavy training stress, mood state, and performance for male junior rowers. Perceptual and Motor Skills, 95, 520-526.

Kallus, K. W. (1995). The Recovery-Stress Questionnaire. Frankfurt: Swets and Zeitlinger.Kallus, K. W., & Kellmann, M. (2000). Burnout in athletes and coaches. In Y. L. Hanin (Ed.),

Emotions in sport (pp. 209-230). Champaign, IL: Human Kinetics.Kellmann, M. (2002). Underrecovery and overtraining: Different concepts - similar impact? In

M. Kellmann (Ed.), Enhancing recovery: Preventing underperformance in athletes (pp. 3-24). Champaign, IL: Human Kinetics.

Kellmann, M., & Kallus, K. W. (2000). Der Erholungs-Belastungsfragebogen für Sportler. Handanweisung [The Recovery-Stress Questionnaire for Athletes. Manual]. Frankfurt: Swets and Zeitlinger.

Kellmann, M., & Kallus, K. W. (2001). Recovery-Stress Questionnaire for Athletes. User manual. Champaign, IL: Human Kinetics.

Lehmann, M., Foster, C., Gastmann, U., Keizer, H. A., & Steinacker, J. M. (1999). Definition, types, symptoms, findings, underlying mechanisms, and frequency of overtraining and overreaching syndrome. In M. Lehmann, C. Foster, U. Gastmann, H. A. Keizer, & J. M. Steinacker (Eds.), Overload, fatigue, performance incompetence, and regeneration in sport (pp. 1-6). New York: Plenum.

McDonald, R. P., & Ringo Ho, M. H. (2002). Principles and practice in reporting structural equation analyses. Psychological Methods, 7, 64-82.

McNair, D., Lorr, M., & Droppleman, L. (1971). Profile of mood states manual. San Diego: Educational and Testing Service.

Mäestu, J., Jürimäe, J., Kreegipuu, K., & Jürimäe, T. (2006). Changes in perceived stress and recovery during heavy training in highly trained rowers. The Sport Psychologist, 20, 24-39.

Martin, D. T., Andersen, M. B., & Gates, W. (2000). Using Profile of Mood States (POMS) to monitor high-intensity training in cyclists: Group versus case studies. The Sport Psychologist, 14, 138-156.

Miké, V. (2003). Evidence and the future of medicine. Evaluation & the Health Professions, 26, 127-152.

Nevitt, J., & Hancock, G. R. (1997). Relative performance of rescaling and resampling approaches to model chi-square and parameter standard error estimation in structural equation modeling. Paper presented at the annual meeting of the American Educational Research Association, San Diego, CA.

Nunnally, J. C. (1978). Psychometric theory. New York: McGraw Hill.Nunnally, J. C., & Bernstein, I. H. (1994). Psychometric theory. New York: McGraw Hill.Prapavessis, H., Grove, J. R., Maddison, R., & Zillmann, N. (2003). Self-handicapping tendencies,

coping and anxiety responses among athletes. Psychology of Sport and Exercise, 4, 357-375.Thomas, R. J., & Nelson, J. K. (2001). Research methods in physical activity. Champaign, IL: Human

Kinetics.


Urhausen, A., & Kindermann, W. (2002). Diagnosis of overtraining: What tools do we have? Sports Medicine, 32, 95-102.

Van de Vijver, F., & Hambleton, R. K. (1996). Translating tests: Some practical guidelines. European Psychologist, 1, 89-99.

Waite, B. T., Gansneder, B., & Rotella, R. J. (1990). A sport-specific measure of self-acceptance. Journal of Sport and Exercise Psychology, 12, 264-279.

West, S. G., Finch, J. F., & Curran, P. J. (1995). Structural equation models with non-normal variables: Problems and remedies. In R. H. Hoyle (Ed.), Structural equation modeling (pp. 56-75). Thousand Oaks, CA: Sage.

Wheaton, B., Muthén, B., Alwin, D. F., & Summers, G. F. (1977). Assessing reliability and stability in panel models. In D. R. Heise (Ed.), Sociological methodology 1977 (pp. 84-136). San Francisco, CA: Jossey-Bass.

spanish adaptation and analysis by structural equation modeling...

Documents