Professional Documents
Culture Documents
Dwyer2012 PDF
Dwyer2012 PDF
To cite this article: John J. M. Dwyer , Tala Chulak , Scott Maitland , Kenneth R. Allison , Daria C. Lysy , Guy E. J. Faulkner & Judy
Sheeshka (2012) Adolescents‘ Self-Efficacy to Overcome Barriers to Physical Activity Scale, Research Quarterly for Exercise and Sport,
83:4, 513-521, DOI: 10.1080/02701367.2012.10599140
Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the
publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or
warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed
by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with
primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings,
demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly
in connection with, in relation to or arising out of the use of the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction,
redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly
forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-
conditions
Dwyer, Chulak, Maitland, Allison,
Measurement
Lysy, Faulkner,
and and
Evaluation
Sheeshka
This paper describes a revised measure of self-efficacy to overcome barriers to moderate and vigorous physical activity in a sample of
Downloaded by [University of Cambridge] at 17:04 19 December 2014
484 high school students in Toronto, Ontario. The students had a mean age of 15.3 years. Principal axis factoring with oblique
rotation yielded five factors: self-efficacy to overcome internal, harassment, physical environment, social environment, and responsi-
bilities barriers. Two problematic items were removed, which resulted in a 22-item measure. Subsequent analyses were conducted on
responses to this shortened measure. Confirmatory factor analysis supported the five-factor model and demonstrated age- and sex-
invariance. The subscales had good internal consistency reliability. Structural regressions demonstrated a strong relationship between
the resulting factors and a physical activity measure (energy expenditure), showing predictive validity.
Key words: factor analysis, validity, reliability & Irving, 2007; Grunbaum et al., 2002). Physical inactivity
during adolescence is a concern considering that physical
al., 2006; Hohepa, Schofield, & Kolt, 2006), having a busy or vigorous physical activity) influenced the relationship
social life (Allison et al., 2005), having friends and parents between self-efficacy and physical activity levels among
who do not support physical activity (Allison et al., 2005; youth. Dwyer et al.’s measure addresses many barriers,
Dwyer et al., 2006), and not having friends to do physi- but subsequent qualitative research among adolescents
cal activity with (Allison et al., 2005). Lacking motivation identified additional barriers the measure should address
(Hohepa et al., 2006), being bored (Allison et al., 2005), (e.g., being teased, bullied, or intimidated by others dur-
the stress of competition (Dwyer et al., 2006), feeling ing physical activity; Allison et al., 2005; Dwyer et al., 2006).
intimidated by others when playing sports (Allison et al., The purpose of our study was to revise the Dwyer et
2005), insufficient physical skills (Hohepa et al., 2006), al. (1998) self-efficacy measure by including additional
concern that peers would tease them for playing poorly barriers identified in previous qualitative research (Allison
(Allison et al., 2005), being self-conscious about their et al., 2005; Dwyer et al., 2006) and using moderate and
physical appearance and being overweight (Dwyer et al., vigorous physical activity (rather than exercise or vigorous
2006), and injury (Allison et al., 2005) are also barriers. physical activity) as a reference point, which is consistent
Additional barriers include unavailable or inaccessible with the recommended mode of physical activity (Centers
facilities and programs (e.g., costly programs, lack of for Disease Control and Prevention, 2008). The revised
transportation to facilities, and staying home because of measure is the Self-Efficacy to Overcome Barriers to Physi-
the weather; Dwyer et al., 2006; Hohepa et al., 2006) and cal Activity Scale (SOBPAS). We examined the factorial
Downloaded by [University of Cambridge] at 17:04 19 December 2014
concern about unsafe facilities (e.g., facilities where gangs validity, internal consistency reliability, and predictive
hang out; Dwyer et al., 2006; Hohepa et al., 2006). validity of the SOBPAS. This measure could be used in
Previous measures of self-efficacy to overcome bar- research examining topics such as correlates of physical
riers to physical activity participation had shortcomings, activity and physical activity intervention research.
in particular poor content validity due to inadequately
addressing some important barriers. The Self-Efficacy
for Exercise Behaviors Scale (Sallis, Pinski, Grossman, Method
Patterson, & Nader, 1988) has items referring to an ex-
ercise program; thus, the content of the items reflects a Participants and Procedure
narrower component of physical activity (i.e., structured Five schools were conveniently selected in Toronto,
physical activity to achieve a cardiovascular effect) rather Ontario, and students in grades 9, 10, and 11 were invited
than physical activity broadly. Also, factor analysis results to participate in the study. University and school board
were weak. The self-efficacy measure in the Reynolds et al. committees provided research ethics approval for the
(1990) study has item content that reflects only exercise research protocol. An information letter and parental con-
and has poor content validity. Also, factor analysis did not sent form (active consent) were sent home with students.
examine construct dimensionality. The Self-Efficacy Scale During February to May 2005, research staff administered
(Saunders et al., 1997) does not address important bar- the self-report measures in the schools to students who
riers, such as being teased and bullied by others, and the returned a signed consent form. Five hundred eighteen
stress of competition and internal consistency reliabilities students completed the measure. We removed 34 par-
of the subscales are low. Ryan and Dzewaltowski (2002) ticipants with outliers in the Self-Administered Physical
did not examine the factorial validity of the Barriers Ef- Activity Checklist (SAPAC) data (more than 3 hr/day on a
ficacy Scale. Item response modeling, which focuses on specific physical activity). This resulted in a sample of 484
item difficulty, was used to develop the Physical Activity participants consisting of 177 students in grade 9, 149 in
Self-Efficacy Scale (Jago et al., 2009). It did not include grade 10, and 158 in grade 11. There were 14- (n = 117),
various barriers, such as teasing and bullying by others, or 15- (n = 162), 16- (n = 147), 17- (n = 55), and 18-year-olds
facilities not being available or accessible. Also, there were (n = 3) in the sample; M age = 15.3 years, SD = 1.0. There
insufficient easy and difficult items; thus, more psychomet- were 194 boys (40%) and 290 girls (60%). This sample
ric research on the measure is needed (Jago et al., 2009). and the outlier group did not significantly differ in score
Dwyer, Allison, and Makin (1998) adapted measures distribution for age (χ2 = 3.48, p ≥ .01) and gender (χ2 =
used in previous research (Reynolds et al., 1990; Sallis et 4.62, p ≥ .01).
al., 1988) to develop their self-efficacy measure. Analysis
yielded two factors: self-efficacy to overcome (a) external
Measures
barriers and (b) internal barriers. As a cautionary note,
this measure assesses self-efficacy to overcome barriers We pilot tested the measures among 90 students in
related to vigorous physical activity rather than both mod- one high school (45 students in grade 9, 28 students in
erate and vigorous physical activity. Ryan and Dzewaltowski grade 10, and 17 students in grade 11) to determine
(2002) found that the type of physical activity referred to whether they had difficulty answering specific questions
in self-efficacy measures (vigorous exercise vs. moderate and to improve measures as necessary. The SOBPAS is a
revision of a 20-item measure of self-efficacy to overcome recall form in the current study to assess physical activity
barriers to vigorous physical activity among high school participation. We used the modified SAPAC to determine
students (Dwyer et al., 1998). The Dwyer et al. measure, metabolic equivalent (MET) hours/week (energy ex-
which consisted of an external barriers subscale (e.g., lack penditure). Participants received a list of 25 activities, such
of support from friends) and an internal barriers subscale as soccer and dance, and five spaces to list other activities.
(e.g., not having fun), had good internal consistency reli- They reported the number of days they performed the
ability and did not correlate significantly with social desir- activities in the previous 7 days and the average number
ability. We made several revisions to the measure based on of minutes they did each activity. The frequency (days),
the findings of qualitative studies on barriers to physical duration (hours), and MET values representing the inten-
activity among adolescents. We removed “feel stressed,” sity of each activity for youth (Ridley, Ainsworth, & Olds,
because the Dwyer et al. measure already included items 2008) were multiplied and summed to determine MET
reflecting sources of stress (e.g., too much schoolwork, hours/week. Other researchers have used this calcula-
having a job, and family responsibilities). We replaced tion procedure (Imperatore, Cheng, Williams, Fulton, &
“lack willpower” and “do not feel in the mood” with “not Gregg, 2006). We calculated only activities with an MET
motivated” (Hohepa et al., 2006) and replaced “long-term value of 3.0 or higher (i.e., moderate to vigorous intensity)
illness, disability, or injury” with the more general “being to be consistent with the SOBPAS focusing on self-efficacy
sick or injured.” We added six additional barriers based to overcome barriers related to moderate or vigorous
Downloaded by [University of Cambridge] at 17:04 19 December 2014
on qualitative studies: “weather” (Allison et al., 2005), physical activity. We considered participants reporting
“not having a safe place to do physical activity” (Dwyer more than 3 hr/day on a specific physical activity to be
et al., 2006), “being bullied or intimidated by others” outliers and removed them. MET hours/week were posi-
and “being teased by friends during physical activity or tively skewed, so a square root transformation was done
sports” (Allison et al., 2005; Hohepa at al., 2006), “not to yield a normal distribution.
having transportation to facilities” (Allison et al., 2005),
and “too much competition” (Allison et al., 2005; Dwyer
Statistical Analyses
et al., 2006). The SOBPAS question stem states: “How
confident are you that you can overcome this thing We used SPSS to perform statistical analyses. Descrip-
and still do moderate or vigorous physical activities?” It tive statistics for demographics were calculated, and Prin-
provided examples of moderate and vigorous physical cipal Axis Factoring (PAF, which focuses on identifying a
activities. The question stem identifies the type of physical theoretical factor solution) (Tabachnick & Fidell, 2007)
activity but does not specify frequency, duration, and time was performed on the SOBPAS items. Pairwise deletion
frame, which is consistent with an approach used previ- was used to deal with missing data.
ously (Dwyer et al., 1998; Saunders et al., 1997). Thus, the We extended our analyses to incorporate confirma-
SOBPAS measures self-efficacy to overcome barriers to tory factor analysis (CFA) to determine the strength of
moderate and vigorous physical activity in general rather the SOBPAS model. Models were tested using Amos
than barriers related to a specific type of physical activity, 17 (SPSS, 2009). We analyzed covariance matrices with
which was inconsistent in previous research. For example, results reported as standardized estimates for ease of in-
Ryan and Dzewaltowski (2002) specified at least 60 min terpretation. For factor scaling, we set one factor loading
of moderate or vigorous physical activity on most days to 1.0 for each factor. The chi-square difference test (Δχ2;
of the week, Jago et al. (2009) used 30 min or more of Joreskog & Sorbom, 1987) was used to compare nested
physical activity either “one day” or at least 4 days/week, models and evaluate whether model modifications pro-
and Sallis et al. (1988) referred to exercise during at least vided significant improvement at each step. Model fit was
the next 6 months. Twenty-four barriers are listed, and evaluated by examining the following fit indexes: model
response options are 1 = not at all confident, 2 = not very χ2 and the ratio of χ2/df (Bollen, 1989), goodness of fit
confident, 3 = somewhat confident, 4 = confident, and index (GFI; Joreskog & Sorbom, 1987), non-normed fit
5 = very confident. Also, participants reported their age, index (NNFI; Bentler & Bonnett, 1980), comparative fit
grade level, and sex and completed other measures not index (CFI; Bentler, 1990), and root mean square error
directly related to the focus of this article. of approximation (RMSEA; Steiger, 1990). The model fit
We used a modified version of the SAPAC to assess well if a consensus of measures met or exceeded gener-
participation in physical activity. Originally, the SAPAC was ally accepted levels. It is widely accepted that model χ2
developed as a 1-day recall of physical activity participa- should be nonsignificant if the model fits well, although
tion and selected sedentary activities (Sallis et al., 1996). this measure is sensitive to large sample sizes. Therefore,
The physical activity component was shown to have good we consulted GFI, NNFI, and CFI values, and all should
reliability and validity (Sallis et al., 1996). The SAPAC has exceed .90 to indicate acceptable fit. Additionally, RMSEA,
been adapted to a 7-day recall of these activities (Marshall, which demonstrates the amount of error variance per
Biddle, Sallis, McKenzie, & Conway, 2002). We used this degree of freedom, should result in values less than .05.
Beyond simply ensuring that a measurement model PAF relates to theory development or identifying
fits the data well, the extension to multigroup factor underlying constructs (Tabachnick & Fidell, 2007); thus,
models tests whether measures mark the same constructs we used this factor extraction technique. It is recom-
across comparison groups (Maitland, Nyberg, Bäckman, mended that researchers explore different numbers of
Nilsson, & Adolfsson, 2009). We conducted tests of mea- factors to identify the best factor solution (Tabachnick &
surement equivalence or invariance (ME/I; Vandenberg Fidell, 2007). We used several methods (eigenvalues > 1,
& Lance, 2000) for age and gender groups (n 14–15 = Kaiser, 1960; scree plot, Cattell, 1966; percent of variance
240, n 16–18 = 162; n boys = 167, n girls = 235). To address explained in extracted factors, the residuals, and factor
whether the factor solution was the same for boys and interpretability) to identify the number of factors. Only
girls, we tested ME/I of the best fitting model. First, the items with factor loadings of .32 or larger, which represents
configural model assessed whether the factor structure 10% variance overlap between an item and factor, were
was the same for boys and girls. Once equivalence was interpreted (Tabachnick & Fidell, 2007). PAF with oblique
established, we tested three further hypotheses: equivalent (direct oblimin) rotation, using Kaiser’s criterion, initially
factor loadings, equivalent relationships between factors, yielded a three-factor solution. We used oblique rotation
and the hypothesis of equivalent measurement residuals. because previous research showed moderate to large cor-
We followed similar procedures to test for invariance relations among self-efficacy subscales (Dwyer et al., 1998;
between the two age groups. We conducted analyses on Saunders et al., 1997). In contrast to Kaiser’s criterion re-
Downloaded by [University of Cambridge] at 17:04 19 December 2014
all participants with complete data, thus allowing the GFI sults, the scree plot to identify distinct changes in the slope
and modification indexes to be calculated. for plotted eigenvalues (Cattell, 1966) indicated it was best
We analyzed internal consistency reliability on the to retain five factors. The three-factor solution was unclear,
subscales. Finally, structural models regressed the physical as items on various factors did not fit together conceptually,
activity measure (MET hours/week) onto the resulting and some loaded significantly on more than one factor.
SOBPAS subscales to further assess our model validity. We Kaiser’s criterion may over- or underestimate the number
performed a square root transformation on MET hours/ of factors (Tabachnick & Fidell, 2007). Thus, we examined
week to normalize the distribution. Evidence of gender two-, four-, five-, and six-factor solutions, with direct obli-
and age invariance controlled for differences accountable min rotation. The two-, four-, and six-factor solutions were
to those personal characteristics. difficult to interpret. Again, the solutions were unclear, as
factor items did not conceptually fit together and some had
significant loadings on more than one factor.
Results The five-factor solution was stable (except for the un-
supportive parent item, there were no double loadings of
items on factors), had fewer residuals with absolute values
Exploratory Factor Analysis (EFA)
greater than .05 (3%), and had interpretable factors (fac-
We examined several assumptions of factor analysis tor items fit conceptually to reflect specific constructs);
(Tabachnick & Fidell, 2007). The sample size of 484 was therefore, we deemed it the best solution. Table 1 shows
above the recommended 300 for a reliable factor analy- factor loadings in the factor pattern matrix, the commu-
sis. Z scores and squared multiple correlations (SMCs) nalities, and the variance explained by the solution. The
showed no univariate and multivariate outliers, respec- communalities indicated that the five factors accounted
tively (Tabachnick & Fidell, 2007). Skewness and kurtosis for 29–74% of the variance in each item, and the solution
values, histograms, and normal Q-Q plots showed nor- accounted for 54% of the variance. In all factor solutions,
mally distributed data (Field, 2009). Scatter plots of items the barrier “weather” had lower factor loadings and less
showed the assumption of linearity was met (Tabachnick fit within a factor. For exploratory purposes, this item was
& Fidell, 2007). Multicollinearity was not present in the removed and all factor solutions were reanalyzed. Remov-
matrix, as correlations were below the .80 cut-off value and ing the item produced factor solutions that were less in-
SMCs were not close to 1.0 (Tabachnick & Fidell, 2007). terpretable; thus, the item was retained in the five-factor
Next, we assessed the factorability of the correlation solution in EFA. The factors were labeled internal barriers,
matrix. There was a relationship between the items in that harassment barriers, physical environment barriers, social
Bartlett’s test of sphericity was significant (χ2 = 5,799.12, p < environment barriers, and responsibilities barriers.
.001), and the Kaiser-Meyer-Olkin value exceeded Kaiser’s
(1974) recommendation of > .90 (Tabachnick & Fidell,
CFA and Gender and Age Invariance
2007). Also, measures of sampling adequacy for each item
were above .93, larger than the recommended value of > The five-factor model was specified in Amos; the
.70 (Kaiser, 1974). Examination of the correlation matrix initial model fit was reasonable but required some modifi-
indicated that all items correlated significantly (p < .05) cations. The weather and unsupportive parent items were
and positively with each other, ranging from .21 to .73. removed, and data for the 22-item SOBPAS were used in
the remaining analyses. Analyzing the model resulted in: for: (a) how confident are you that you can overcome
χ2 Baseline = 611.81, df = 199, χ2/df = 3.07, p < .001, GFI community programs and facilities not being available,
=.87, NNFI = .90, CFI = .91, RMSEA = .07. Results were and (b) how confident are you that you can overcome
mixed, as the NNFI and CFI were good, whereas other fit school programs not being available. Reanalyzing the
indexes suggested model improvement was warranted. model with these modifications resulted in improved fit:
Modification indexes suggested correlating residuals for: χ2 M1 = 548.00, df = 197, χ2/df = 2.78, p < .001, GFI =.89,
(a) how confident are you that you can overcome feeling NNFI = .92, CFI = .93, RMSEA = .07. The NNFI, CFI, and
tired and low on energy, and (b) how confident are you GFI values improved. The GFI value did not exceed .90
that you can overcome not feeling motivated. It was also but improved nonetheless. Although the modifications
apparent that residuals should be allowed to correlate did not improve the RMSEA value, they did not make
this index worse. One would not necessarily expect every
fit index to improve when making modifications that are
not extremely substantial. Our global assessment showed
Table 1. Principal axis factoring with direct oblimin rotation on an overall pattern indicating the modifications improved
SOBPAS items
the model. Comparing the two models showed a signifi-
Self-efficacy item (a–x) Factor Com. cant improvement: Δχ2 Baseline–M1 (2) = 63.81, p < .001;
1 2 3 4 5 therefore, we accepted the second model as the best fit.
Downloaded by [University of Cambridge] at 17:04 19 December 2014
also accepted: χ2 M7 = 865.89, df = 411, χ2/df = 2.11, p accepted model of invariance of factor variances/covari-
< .001, GFI = .84, NNFI = .90, CFI = .91, RMSEA = .05. ances. This result was also accepted as plausible at p < .01:
Comparison of the models supported the hypothesis of χ2 M9 = 925.23, df = 450, χ2/df = 2.06, p < .001, GFI = .82,
equivalent factor loadings: Δχ2 M7–M6 (17) = 19.36, p = NNFI = .90, CFI = .90, RMSEA = .05, Δχ2 M9–M8 (24) =
.308. The next comparison examined whether the factor 41.64, p = .014. The standardized solution demonstrating
variances and covariances could be constrained between invariance of all coefficients including residuals is shown
age groups: χ2 M8 = 883.61, df = 426, χ2/df = 2.07, p < .001, in Figure 1.
GFI = .83, NNFI = .90, CFI = .91, RMSEA = .05. Comparing
the measurement weights model with the more restric-
Internal Consistency Reliability
tive model constraining factor variances/covariances
demonstrated acceptable invariance: Δχ2 M8–M7 (15) = Cronbach’s coefficient alpha for the SOBPAS sub-
17.72, p = .278. The final test of invariance constrained the scales were: self-efficacy to overcome (a) internal barriers,
measurement residuals to be equal between the two age .86, (b) harassment barriers, .81, (c) physical environment
groups and compared this model against the previously barriers, .82, (d) social environment barriers, .80, and (e)
Downloaded by [University of Cambridge] at 17:04 19 December 2014
Figure 1. Standardized solution for final Self-Efficacy to Overcome Barriers to Physical Activity Scale (SOBPAS) model. The
values next to each observed measure are squared multiple correlations, which indicate how much variance in each observed
measure is accounted for by the factors.
responsibilities barriers, .79. These values were above the 1998; Saunders et al., 1997). However, there are mixed
suggested .70 value (Nunnally, 1978), indicating good views about the dimensionality of self-efficacy for physi-
internal consistency reliability. cal activity. For example, using CFA, Motl et al. (2000)
suggested that self-efficacy to overcome physical activity
Predictive Validity of SOBPAS barriers is unidimensional. However, they examined the
factorial validity among adolescent girls only, and their
We used structural regressions to examine the rela- results cannot be generalized to adolescent boys.
tionship between the five SOBPAS latent factors and the In our study, the five factors logically describe the
physical activity measure (MET hours/week). The model factor content. The first subscale addressed self-efficacy to
fit declined from the measurement model described: χ2 overcome internal barriers such as embarrassment about
M10 = 1341.38, df = 220, χ2/df = 6.09, p < .001, GFI = .82, others watching, concern about weight, insufficient skills,
NNFI = .74, CFI = .77, RMSEA = .11. The structural regres- not feeling motivated, and having other interests. The sec-
sions between the SOBPAS latent factors and physical ond subscale dealt with being teased and bullied by others
activity level were all statistically significant and ranged and was named self-efficacy to overcome harassment barri-
from β = .09 (responsibilities barriers) to β = .25 (internal ers. Participants in qualitative research expressed concerns
barriers). Overall, the five SOBPAS factors accounted for about safety by referring to gangs hanging out at physical
44% of the variance in physical activity level (see Table 2). activity facilities and crime at facilities (Dwyer et al., 2006;
Downloaded by [University of Cambridge] at 17:04 19 December 2014
First, the SOBPAS measures self-efficacy to overcome bar- tive research (Allison et al., 2005; Dwyer et al., 2006),
riers to physical activity, whereas some others focus on self- and unlike other instruments it measures self-efficacy to
efficacy to overcome impediments to exercise (Reynolds et overcome barriers to both moderate and vigorous physical
al., 1990; Sallis et al., 1988). Exercise is a structured physical activity. Analyses of SOBPAS data from a sample of male
activity that is done for fitness reasons; thus, self-efficacy and female high school students in different grades in-
to overcome barriers to exercise and physical activity in dicated the SOBPAS is a multidimensional measure of
general may be different. Second, the SOBPAS measures self-efficacy to overcome barriers and has good construct
self-efficacy to overcome barriers to both moderate and validity, internal consistency reliability, and predictive
vigorous physical activity, whereas the Dwyer et al. measure validity. Results suggest the SOBPAS shows promise in
deals with self-efficacy to overcome barriers to vigorous research on the correlates of physical activity and physi-
physical activity only. Some people may be less confident cal activity intervention research targeting self-efficacy to
that they can handle impediments, such as not having fun overcome barriers as a determinant of behavior change.
and not feeling motivated, in more strenuous physical
activities. Third, the number of items in various measures
likely influenced the factor structure; including more items References
allows various dimensions to be examined more compre-
Allison, K. R., Adlaf, E. M., Dwyer, J. J. M., Lysy, D. C., & Irving,
hensively. For example, the SOBPAS assesses self-efficacy
Downloaded by [University of Cambridge] at 17:04 19 December 2014
Grunbaum, J. A., Kann, L., Kinchen, S. A., Williams, B., Ross, Ries, A. V., Gittelsohn, J., Voorhees, C. C., Roche, K. M., Clifton,
J. G., Lowry, R., & Kolbe, L. (2002). Youth risk behavior K. J., & Astone, N. M. (2008). The environment and urban
surveillance—United States, 2001. Journal of School Health, adolescents’ use of recreational facilities for physical activ-
72, 313–328. ity: A qualitative study. American Journal of Health Promotion,
Hohepa, M., Schofield, G., & Kolt, G. S. (2006). Physical activity: 23(1), 43–50.
What do high school students think? Journal of Adolescent Ryan, G. J., & Dzewaltowski, D. A. (2002). Comparing the
Health, 39, 328–336. relationships between different types of self-efficacy and
Imperatore, G., Cheng, Y. J., Williams, D. E., Fulton, J., & Gregg, physical activity in youth. Health Education and Behavior,
E. W. (2006). Physical activity, cardiovascular fitness, and 29, 491–504.
insulin sensitivity among U.S. adolescents: The National Sallis, J. F., Pinski, R. B., Grossman, R. M., Patterson, T. L., &
Health and Nutrition Examination Survey, 1999–2002. Nader, P. R. (1988). The development of self-efficacy scales
Diabetes Care, 29, 1567–1572. for health-related diet and exercise behaviors. Health Educa-
Jago, R., Baranowski, T., Watson, K., Bachman, C., Baranowski, tion Research, 3, 283–292.
J. C., Thompson, D.,...Moe, E. (2009, March 31). Devel- Sallis, J. F., Strikmiller, P. K., Harsha, D. W., Feldman, H. A.,
opment of new physical activity and sedentary behavior Ehlinger, S., Stone, E. J.,...Woods, S. (1996). Validation of
change self-efficacy questionnaires using item response interviewer- and self-administered physical activity check-
modeling. International Journal of Behavioral Nutrition and lists for fifth grade students. Medicine & Science in Sports &
Physical Activity, 6, 20. Exercise, 28, 840–851.
Downloaded by [University of Cambridge] at 17:04 19 December 2014
Joreskog, K. G., & Sorbom, D. (1987). LISREL 7. A guide to the Saunders, R. P., Pate, R. R., Felton, G., Dowda, M., Weinrich,
program and applications. Chicago: SPSS. M. C., Ward, D. S.,...Baranowski, T. (1997). Development
Kaiser, H. F. (1960). The application of electronic computers of questionnaires to measure psychosocial influences on
to factor analysis. Educational and Psychological Measurement, children’s physical activity. Preventive Medicine, 26, 241–247.
20, 141–151. SPSS. (2009). Amos 17.0 [Computer software]. Chicago: SPSS.
Kaiser, H. F. (1974). An index of factorial simplicity. Psychometrika, Steiger, J. H. (1990). Structural model evaluation and modifica-
39(1), 31–36. tion: An interval estimation approach. Multivariate Behav-
Loucaides, C. A., Plotnikoff, R. C., & Bercovitz, K. (2007). ioral Research, 25, 173–180.
Differences in the correlates of physical activity between Strong, W. B., Malina, R. M., Blimkie, C. J., Daniels, S. R., Dish-
urban and rural Canadian youth. Journal of School Health, man, R. K., Gutin, B.,...Trudeau, F. (2005). Evidence based
77, 164–170. physical activity for school-age youth. Journal of Pediatrics,
Maitland, S. B., Nyberg, L., Bäckman, L., Nilsson, L-G., & Adolfs- 146, 732–737.
son, R. (2009). On the structure of personality: Are there Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statis-
separate temperament and character factors? Personality tics (5th ed.). Boston: Allyn and Bacon.
and Individual Differences, 47, 180–184. Telama, R., Yang, X., Viikari, J., Valimaki, I., Wanne, O., &
Marshall, S. J., Biddle, S. J. H., Sallis, J. F., McKenzie, T. L., & Raitakari, O. (2005). Physical activity from childhood to
Conway, T. L. (2002). Clustering of sedentary behaviors adulthood: A 21-year tracking study. American Journal of
and physical activity among youth: A cross-national study. Preventive Medicine, 28, 267–273.
Pediatric Exercise Science, 14, 401–417. Vandenberg, R. J., & Lance, C. E. (2000). A review and synthesis
Motl, R. W., Dishman, R. K., Trost, S. G., Saunders, R. P., Dowda, of the measurement invariance literature: Suggestions,
M., Felton, G.,...Pate, R. R. (2000). Factorial validity and practices, and recommendations for organizational re-
invariance of questionnaires measuring social-cognitive search. Organizational Research Methods, 3(1), 4–70.
determinants of physical activity among adolescent girls. Van Der Horst, K., Paw, M. J., Twisk, J. W., & Van Mechelen,
Preventive Medicine, 31, 584–594. W. (2007). A brief review on correlates of physical activity
Noland, H., Price, J. H., Dake, J., & Telljohann, S. K. (2009). and sedentariness in youth. Medicine & Science in Sports &
Adolescents’ sleep behaviors and perceptions of sleep. Exercise, 39, 1241–1250.
Journal of School Health, 79, 224–230.
Nunnally, J. C. (1978). Psychometric theory (2nd ed.). New York:
McGraw-Hill. Authors’ Notes
Petosa, R. L., Hortz, B. V., Cardina, C. E., & Suminski, R. R.
(2005). Social cognitive theory variables associated with
This research was funded by the Heart and Stroke Founda-
physical activity among high school students. International
Journal of Sports Medicine, 26, 158–163.
tion of Ontario (HBR4860). Please address correspond-
Reynolds, K. D., Killen, J. D., Bryson, S. W., Maron, D. J., Taylor, ence concerning this article and requests for a copy of the
C. B., Maccoby, N., & Farquhar, J. W. (1990). Psychosocial SOBPAS to John Dwyer, Department of Family Relations
predictors of physical activity in adolescents. Preventive and Applied Nutrition, College of Social and Applied Hu-
Medicine, 19, 541–551. man Sciences, University of Guelph, Macdonald Institute
Ridley, K., Ainsworth, B. E., & Olds, T. S. (2008). Development Building, 50 Stone Road East, Guelph, Ontario, Canada,
of a compendium of energy expenditures for youth. Inter- N1G 2W1.
national Journal of Behavioral Nutrition and Physical Activity,
5, 45–52. E-mail: dwyer@uoguelph.ca