درباره ماهیت و عملکرد پروتکل های امتیاز دهی مورد استفاده در تحقیقات انگیزه ورزش: مطالعه تجربی از مقررات رفتاری در پرسشنامه ورزش

Abstract Objectives The purpose of this study was to examine the effects of different scoring protocols used with instruments designed to assess motivation in line with Organismic Integration Theory (OIT; Deci & Ryan, 2002). Design This study used non-probability based sampling within a cross-sectional (survey) design. Methods Participants across four samples (N's ranged from 236 to 1200) completed either (a) the Behavioral Regulation in Exercise Questionnaire (BREQ), (b) the BREQ-2, or (c) the BREQ-2R in conjunction with a self-report assessment of physical activity behavior

.

Results Preliminary analyses: missing data and assumptions of normality No out-of-range responses were observed in the data. Missing data was evident across all samples except sample 1. Little's (1988) test indicated that the missing data noted in sample 2 (Little's χ2 = 390.69, df = 359, p = .12) could be considered missing completely at random, while the missing data in sample 4 (Little's χ2 = 204.56, df = 145, p < .01) could at best be considered missing at random. All missing data were replaced with estimated values using an expectation maximization algorithm. A portion of the data provided in sample 2 (n = 30–39), sample 3 (n = 6) and sample 4 (n = 3) was discarded given the presence of large standardized residuals (>|3.00| SD's) linked with LTEQ scores. Additional cases were removed on a model-by-model basis in the multiple regression analyses based on large standardized residuals (>|3.00|) for the model under scrutiny. Estimates of internal consistency score reliability Greater variability was evident in coefficient-α's (see Tables 2 and 3) reported for the item-aggregation approach (Mα = 0.81, SDα = 0.12) compared to the bifurcation approach (Mα = 0.84, SDα = 0.07). This is partly due to the value for coefficient-α recorded in sample 4 for responses to the BREQ-2's amotivation subscale. Estimates of score reliability for autonomous motivation (Mα = 0.89, SDα = 0.03) were higher than controlling motivation (Mα = 0.78, SDα = 0.04; see Table 3). Table 2. Descriptive statistics and estimates of internal consistency reliability for all constructs derived from the item-aggregation scoring approach. Samples Variables M SD Skew. Kurt. α 1 1. External regulation 0.63 0.74 1.37 1.82 0.84 2. Introjected regulation 2.15 1.01 −0.10 −0.48 0.79 3. Identified regulation 3.30 0.62 −0.70 −0.27 0.71 4. Intrinsic regulation 3.12 0.85 −0.95 0.33 0.91 5. Physical activity 52.24 18.70 0.20 −0.47 – 2 1. External regulation 0.65 0.82 1.55 2.30 0.83 2. Introjected regulation 1.60 1.13 0.37 −0.81 0.80 3. Identified regulation 2.71 0.98 −0.51 −0.55 0.84 4. Intrinsic regulation 2.30 1.20 −0.24 −1.05 0.94 5. Physical activity 35.12 24.34 1.81 7.86 – 3 1. Amotivation 0.21 0.38 1.82 2.39 0.81 2. External regulation 1.43 0.55 1.53 2.36 0.82 3. Introjected regulation 2.22 0.85 0.38 −0.75 0.80 4. Identified regulation 3.02 0.68 −0.59 −0.21 0.76 5. Intrinsic regulation 3.04 0.81 −0.54 −0.62 0.91 6. Physical activity 46.90 26.32 0.77 0.13 – 4 1. Amotivation 0.03 0.10 3.53 11.96 0.39 2. External regulation 0.90 0.79 0.89 0.24 0.85 3. Introjected regulation 2.03 1.03 −0.12 −0.62 0.84 4. Identified regulation 3.38 0.57 −1.37 2.97 0.72 5. Integrated regulation 3.14 0.77 −0.95 0.65 0.86 6. Intrinsic regulation 3.23 0.72 −1.45 3.29 0.90 7. Physical activity 68.59 29.73 1.19 3.29 – Note. M = Univariate Mean. SD = Standard Deviation. Skew. = Univariate Skewness. Kurt. = Univariate Kurtosis. α = Coefficient of Internal Consistency Reliability ( Cronbach, 1951). Table options Table 3. Descriptive statistics and estimates of internal consistency reliability for bifurcation and RAI approaches to instrument scoring. Samples Variables M SD Skew. Kurt. α 1 1. Autonomous motives 3.21 0.65 −0.73 −0.04 0.86 2. Controlled motives 1.28 0.68 0.58 0.75 0.77 3. RAI – Version A 6.12 3.02 −0.44 −0.17 – 2 1. Autonomous motives 2.50 1.02 −0.34 −0.82 0.93 2. Controlled motives 1.06 0.76 0.85 0.68 0.78 3. RAI – Version A 4.39 3.84 −0.41 −0.10 – 3 1. Autonomous motives 3.03 0.68 −0.54 −0.46 0.89 2. Controlled motives 1.82 0.55 0.62 0.13 0.74 3. RAI – Version A 4.02 2.53 −0.74 0.78 – 4. RAI – Version B 9.45 4.32 −0.80 0.49 – 4 1. Autonomous motives 3.26 0.58 −0.95 1.43 0.90 2. Controlled motives 1.35 0.77 0.22 −0.33 0.84 3. RAI – Version A 3.15 2.88 −0.38 0.23 – 4. RAI – Version B 12.65 3.89 −0.58 0.57 – 5. RAI – Version C 15.53 4.36 −0.57 0.23 – Note. M = Univariate Mean. SD = Standard Deviation. Skew. = Univariate Skewness. Kurt. = Univariate Kurtosis. α = Coefficient of Internal Consistency Reliability ( Cronbach, 1951). RAI – Version A = Relative Autonomy Index calculated using the weighted subscale procedure advocated by Mullan et al. (1997) in the following formula: Σ[(external * −2) + (introjected * −1) + (identified * 1) + (intrinsic * 2)]. RAI – Version B = Relative Autonomy Index calculated using the weighted subscale procedure proposed by Markland (2011) to include amotivation using the following formula: Σ[(amotivation * −3) + (external * −2) + (introjected * −1) + (identified * 2) + (intrinsic * 3)]. RAI – Version C = Relative Autonomy Index calculated using the protocol extrapolated from the work of Vallerand et al. (2008) according to the following formula: Σ[(amotivation * −3) + (external * −2) + (introjected * −1) + (identified * 1) + (integrated * 2) + (intrinsic * 3)]. Table options Descriptive statistics and bivariate correlations Inspection of the descriptive statistics (see Tables 2and 3) revealed at least two noteworthy observations. First, autonomous motives were consistently more strongly endorsed as reasons for physical activity participation in contrast to controlling motives. Second, amotivation was always the construct least endorsed by any sample followed closely by external regulation in this investigation. Pearson correlations testing bivariate associations between the motives and physical activity behavior using all scoring protocols are presented in Tables A and B (see Supplementary document). A few summary observations are noteworthy. First, support for the hypothesized quasi-simplex pattern of associations between OIT motives was evident aside from sample 4 where the correlation between introjected-integrated regulations marginally exceeded that noted between introjected-identified regulations. Second, identified regulation was the motive most strongly associated with physical activity behavior followed closely by intrinsic and integrated regulations. Third, autonomous motivation was more strongly correlated with more frequent physical activity behavior than controlling motivation. Fourth, the pattern of correlations between autonomous and controlled motives varied in terms of strength and direction per sample but was largely weak in nature (Mr12 = 0.10, SDr12 = 0.12; Range = −0.03–0.25). Finally, higher RAI scores were consistently associated with more frequent physical activity behavior yet the magnitude varied across samples and RAI scoring protocols (Mr12 = 0.22, SDr12 = 0.14; Range = 0.07–0.44). Multiple regression analyses predicting leisure time physical activity from exercise motivation The following section plus Table 4 summarizes the results of the multiple regression analyses.2 Table 4. Summary results from the regression models predicting LTPA from different scoring models. Variables Sample 1 Sample 2 Sample 3 Sample 4 β rs β rs β rs β rs Scoring model 1 Amotivation – – – – 0.11 −0.30 0.25 0.43 External regulation 0.08 −0.03 −0.03 −0.22 −0.08 −0.23 −0.12 −0.10 Introjected regulation −0.05 0.19 0.01 0.36 0.10 0.58 0.06 0.38 Identified regulation 0.33 0.97 0.39 0.98 0.24 0.91 0.27 0.79 Integrated regulation – – – – – – 0.10 0.67 Intrinsic regulation 0.09 0.69 0.19 0.89 0.21 0.86 0.01 0.41 Scoring model 2 Autonomous motives 0.33 0.97 0.55 1.00 0.40 1.00 0.29 1.00 Controlled motives 0.05 0.12 −0.01 0.09 0.05 0.36 0.06 0.27 Scoring model 3 RAI – Version A 0.19 1.00 0.44 1.00 0.29 1.00 0.08 1.00 RAI – Version B – – – – 0.34 1.00 0.07 1.00 RAI – Version C – – – – – – 0.11 1.00 Note. RAI – Version A = Relative Autonomy Index calculated using the weighted subscale procedure advocated by Mullan et al. (1997) in the following formula: Σ[(external * −2) + (introjected * −1) + (identified * 1) + (intrinsic * 2)]. RAI – Version B = Relative Autonomy Index calculated using the weighted subscale procedure proposed by Markland (2011) to include amotivation using the following formula: Σ[(amotivation * −3) + (external * −2) + (introjected * −1) + (identified * 2) + (intrinsic * 3)]. RAI – Version C = Relative Autonomy Index calculated using the weighted subscale procedure advocated by Vallerand et al. (2008) using the following formula: Σ[(amotivation * −3) + (external * −2) + (introjected * −1) + (identified * 1) + (integrated * 2) + (intrinsic * 3)]. Scoring Model 1 = Point estimate subscale scores for BREQ/BREQ-2/BREQ-2R responses across samples in the study. Scoring Model 2 = Global autonomous/controlled motives derived from unweighted BREQ item scores (excluding amotivation) across samples in the study. Scoring Model 3 = Overall RAI score derived from weighted BREQ/BREQ-2/BREQ-2R subscale scores across samples in this study. All standardized beta-coefficients were statistically significant at p < .05 if they exceeded (a) |0.10| in sample 1, (b) |0.15| in sample 2, (c) |0.15| in sample 3, and (d) |0.13| in sample 4. Table options Sample 1: Regression models for different scoring configurations were all statistically significant (p < .05) based on the observed values for the item-aggregation approach (F4,261 = 9.91; p < .01; R = 0.36; View the MathML sourceRadj.2 = 0.12), the bifurcation approach (F2,263 = 16.68; p < .01; R = 0.34; View the MathML sourceRadj.2 = 0.11), and the RAI approach (F1,264 = 9.94; p < .01; R = 0.19; View the MathML sourceRadj.2 = 0.03). Sample 2: Regression models for different scoring configurations tested were all statistically significant (p < .05) based on the observed values for the item-aggregation approach (F4,1155 = 126.20; p < .01; R = 0.55; View the MathML sourceRadj.2 = 0.30), the bifurcation approach (F2,1155 = 245.79; p < .01; R = 0.55; View the MathML sourceRadj.2 = 0.30), and the RAI approach (F1,1157 = 284.29; p < .01; R = 0.44; View the MathML sourceRadj.2 = 0.20). Sample 3: Regression models for different scoring configurations were all statistically significant (p < .05) based on the observed values for the item-aggregation approach (F5,359 = 16.70; p < .01; R = 0.43; View the MathML sourceRadj.2 = 0.18), the bifurcation approach (F2,362 = 38.15; p < .01; R = 0.42; View the MathML sourceRadj.2 = 0.17), and the RAI scoring approaches (RAI-Version A: F1,363 = 35.23; p < .01; R = 0.29; View the MathML sourceRadj.2 = 0.09; RAI-Version B: F1,363 = 47.02; p < .01; R = 0.34; View the MathML sourceRadj.2 = 0.11). Sample 4: Summary observations noted in Table 4 indicate the item-aggregation approach (F6,213 = 7.39; p < .01; R = 0.42; View the MathML sourceRadj.2 < 0.15), bifurcation approach (F2,218 = 11.05; p < .01; R = 0.30; View the MathML sourceRadj.2 = 0.08), and RAI-Version C in Table 4 (F1,220 = 2.79; p = .10; R = 0.11; View the MathML sourceRadj.2 = 0.01) appeared tenable in this sample. Two alternative configurations of the RAI approach resulted in a non-significant regression models in this sample (RAI-Version A: F1,220 = 1.34; p = .25; R = 0.08; View the MathML sourceRadj.2 < 0.01; RAI-Version B: F1,220 = 1.01; p = .32; R = 0.07; View the MathML sourceRadj.2 < 0.01).