Appetite 63 (2013) 4858

Contents lists available at SciVerse ScienceDirect

Appetite
journal homepage: www.elsevier.com/locate/appet

Research report

Self-efcacy for healthy eating and peer support for unhealthy eating
are associated with adolescents food intake patterns q
Amanda Fitzgerald a,, Caroline Heary b, Colette Kelly c, Elizabeth Nixon d, Mark Shevlin e
a
School of Psychology, University College Dublin, Dublin, Ireland
b
School of Psychology, National University of Ireland Galway, Ireland
c
Discipline of Health Promotion, School of Health Sciences, National University of Ireland Galway, Ireland
d
School of Psychology, Trinity College Dublin, Dublin, Ireland
e
School of Psychology and Psychology Research Institute, University of Ulster, Magee Campus, Londonderry, Ireland

a r t i c l e i n f o a b s t r a c t

Article history: Adolescence, with its change in dietary habits, is likely to be a vulnerable period in the onset of obesity. It
Received 26 January 2012 is considered that peers have an important role to play on adolescents diet, however, limited research
Received in revised form 1 December 2012 has examined the role of peers in this context. This study examined the relationship between self-efcacy
Accepted 12 December 2012
for healthy eating, parent and peer support for healthy and unhealthy eating and food intake patterns.
Available online 23 December 2012
Participants were 264 boys and 219 girls (N = 483), aged 1318 years, recruited from post-primary
schools in Ireland. Self-report measures assessed self-efcacy, parent and peer support for healthy eating,
Keywords:
and for unhealthy eating. Dietary pattern analysis, a popular alternative to traditional methods used in
Adolescents
Gender
nutritional research, was conducted on a FFQ to derive food intake patterns. Two patterns were identied
Self-Efcacy labelled healthy food intake and unhealthy food intake. Multi-group modelling was used to evaluate
Social support whether the hypothesized model of factors related to dietary patterns differed by gender. The multi-
Structural equation modelling group model t the data well, with only one path shown to differ by gender. Lower self-efcacy for
Dietary Pattern Analysis healthy eating and higher peer support for unhealthy eating were associated with unhealthy food intake.
Higher self-efcacy was associated with healthy food intake. Prevention programs that target self-ef-
cacy for eating and peer support for unhealthy eating may be benecial in improving dietary choices
among adolescents.
2012 Elsevier Ltd. All rights reserved.

Introduction inuence and mediators to address poor eating behaviours (Gentile

et al., 2009; Gould, Jasik, Lustig, & Garber, 2009).
The quality of adolescents diets in the Western world is of Despite greater nutrient needs during adolescence, there is evi-
increasing concern (British Medical Association [BMA], 2003) with dence that adolescents diet quality declines as a greater proportion
obesity in adolescence now at epidemic proportions (Hedley et al., of meals and snacks are consumed away from the home (Lien, Lytle,
2004; Lobstein, James, & Cole, 2003). Research indicates that the & Klepp, 2001; Neumark-Sztainer, Story, Hannan, & Croll, 2002).
development of effective nutrition strategies requires an under- Adolescents have been shown to consume high intakes of foods rich
standing of the multiple interacting factors that inuence adoles- in fat, sugar, and salt and low intakes of fruits, vegetables, whole-
cents dietary behaviour (McClain, Chappuis, Nguyen-Rodriguez, grains, and calcium-rich foods (Institute of Medicine, 2007). Fur-
Yaroch, & Spruijt-Metz, 2009). There has been a recent move to de- thermore, dietary behaviours established during adolescence tend
velop effective dietary interventions that target multiple levels of to persist into adulthood (Mikkila, Rasanen, Raitakari, Pietnen, &
Vikari, 2005) thus, targeting this age group offers an important
opportunity to inuence long-term dietary behaviours.
q
Socio-Cognitive Theory (SCT; Bandura, 1986) is the most
Statement of research: Using Structural Equation Modelling, this research set out
commonly used theory in interventions to promote healthy
to examine the relationship between perceived self-efcacy and social support for
eating with healthy (i.e., fruit, vegetables, brown bread, low-energy drinks) and eating among adolescents (Cerin, Barnett, & Baranowski, 2009;
unhealthy (i.e., sweets, biscuits, desserts, snacks) food intake patterns among Rinderknecht & Smith, 2004). SCT emphasises the importance of
adolescents. personal, socio-environmental and behavioural factors and the
Corresponding author.
interaction between these factors in inuencing behaviour. This
E-mail addresses: amanda.tzgerald@ucd.ie (A. Fitzgerald), caroline.heary@
interaction is referred to as reciprocal determinism as each factor
nuigalway.ie (C. Heary), colette.kelly@nuigalway.ie (C. Kelly), enixon@tcd.ie
(E. Nixon), m.shevlin@ulster.ac.uk (M. Shevlin). may affect or be affected by the others. Self-efcacy, dened as an

0195-6663/$ - see front matter 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.appet.2012.12.011
 A. Fitzgerald et al. / Appetite 63 (2013) 4858 49

individuals belief in their ability to perform a specic behaviour, is differently by the socioeconomic circumstances of their commu-
considered to be a central determinant of behaviour in SCT. Dietary nity, their ethnicity, and by gender norms and values. In recent
self-efcacy refers to the perceived capability to make healthy food years, it has been recommended that public health strategies to
choices, even when faced with potential barriers (Lubans et al., prevent and reduce obesity should be gender specic (Groth, Fagt,
2012). Research with young people guided by SCT, has shown that Stockmarr, Matthiessen, & Biltoft-Jensen, 2009).
higher dietary self-efcacy is associated with higher intakes of
fruit and vegetables (Bere & Klepp, 2004), calcium-rich foods Patterns of dietary intake
(Ievers-Landis et al., 2003) and lower intakes of fat (Frenn, Malin,
& Bansal, 2003). The majority of studies examining factors related to adoles-
SCT also proposes that socio-environmental factors, such as so- cents dietary intake have traditionally focused on intake of indi-
cial support, can inuence behaviour, both directly and indirectly vidual foods, food groups or nutrients (Cutler, Flood, Hannan,
via personal factors such as self-efcacy. Slavin, & Neumark-Sztainer, 2011). However, a limitation of this
Previous studies have shown that adolescents perceptions of approach is that people do not eat isolated foods or nutrients
parent support for healthy eating was associated with fewer snacks and instead eat meals consisting of a variety of foods with complex
and more fruit, vegetables, bre- and calcium-rich foods (Ayala combinations of nutrients which are likely to be interactive (Hu,
et al., 2007; Stanton, Green, & Fries, 2007; Wind et al., 2006). 2002; Tucker, 2010). The single food and nutrient approach may
Dietary self-efcacy beliefs were also shown to partially mediate be insufcient for taking into account complex interactions among
the relationship between parent support and healthy food intake foods and nutrients in studies of people living in their everyday
among children and adolescents (Ievers-Landis et al., 2003; Young, context. Recently, dietary pattern analysis has emerged as an alter-
Fors, & Hayes, 2004), but did not mediate the relationship between native and complementary approach to the single food and nutri-
peer support and intake in adolescents (Ievers-Landis et al., 2003). ent approach. Dietary pattern analysis considers how multiple
In relation to peer support for healthy eating, there is inconsis- foods, drinks and nutrients are consumed in combination and cre-
tency in the literature. Some studies found that self-reported peer ates dietary variables that more realistically resemble actual die-
support for healthy eating was associated with adolescents tary behaviour (Cutler et al., 2011). The collinerity of foods and
healthy food intake (Larson et al., 2009; Stanton et al., 2007), nutrients can be considered an advantage in dietary pattern anal-
whereas, other studies reported no associations (Finnerty, Reeves, ysis because patterns are created on the basis of usual food con-
Dabinett, Jeanes, & Vgele, 2009; Ievers-Landis et al., 2003). sumption (Tucker, 2010).
Interestingly, qualitative research suggests that peers may Dietary patterns can be derived through data-driven methods
encourage adolescents to consume unhealthy foods (e.g., Croll, such as factor analysis using dietary data collected from a specic
Neumark-Sztainer, & Story, 2001; Watt & Sheiham, 1997), how- study population. Factor analysis uses information reported on die-
ever, little research has quantitatively examined how peer factors tary tools such as food frequency questionnaires (FFQs) to identify
may be related to adolescents unhealthy food intake. common underlying dimensions (patterns) of food consumption. It
In the social support literature, there is a tendency for peer sup- aggregates specic food items (or groups) on the basis of the de-
port to increase during adolescence at the same time as a decrease gree to which food items in the dataset are correlated with one an-
in support from parents (Cheng & Chan, 2004). However, there is other. A summary score for each pattern is then derived and the
also evidence for gender differences in adolescent perceptions individual obtains a score for each dietary pattern identied. Die-
and use of social support (Cheng & Chan, 2004; Colarossi, 2001; tary pattern research has great potential for use in nutrition policy,
Eschenbeck, Kohlmann, & Lohaus, 2007). Adolescent girls perceive especially as it demonstrates the importance of total diet in health
higher levels of peer support than boys, whereas, on the contrary, promotion. Dietary pattern analysis has been used successfully in
adolescent boys perceive higher levels of parent support than girls adult populations to examine associations between socio-demo-
(Cheng & Chan, 2004; Colarossi, 2001). These studies highlight the graphic and lifestyles factors and dietary patterns (Engeset, Alsak-
importance of perceived parental support for boys and perceived er, Ciampi, & Lund, 2005; Sanchez-Villegas, Delgado-Rodriguez,
peer support for girls. In the dietary behaviour literature, studies Martinez-Gonzalez, Irala-Estevez, & J. for the SUN group, 2003),
have shown that parent support for healthy eating is associated and further exploration in adolescent populations is warranted
with healthy food intake among adolescent boys but not girls (Bau- (Cutler et al., 2011).
er, Larson, Nelson, Story, & Neumark-Sztainer, 2009; Larson, Story,
Wall, & Neumark-Sztainer, 2006).
The present study
Gender differences in adolescents dietary intake are widely
noted. Adolescent boys report consuming lower intakes of fruits
Using self-reported variables, this study tested whether self-
and vegetables and higher intakes of fast food and soft drinks than
efcacy mediates the relationship between social support variables
girls (e.g., Lien, Jacobs, & Klepp, 2002). In recent years, research has
(i.e., peer support for healthy eating, peer support for unhealthy
shown that hypothesized predictors of dietary intake may also
eating, parent support for healthy eating, parent support for un-
differ by gender (Bauer et al., 2009; Hanson, Neumark-Sztainer,
healthy eating) and dietary intake patterns in adolescent boys
Eisenberg, Story, & Wall, 2005; Larson et al., 2006, 2008; Zabinski
and girls. These predictions were derived from SCT. In addition,
et al., 2006). For example, Bauer et al. found that maternal and peer
multi-group analyses were conducted to test the hypothesis that
healthy eating concerns and maternal support for healthy eating
gender moderates the associations between self-efcacy and social
were predictive of lower fast-food intake for boys but not girls.
support variables and dietary intake patterns.
Larson et al. (2006) found that self-efcacy was associated with
calcium intake for girls but not boys. Thus, the relationship
between hypothesized factors (such as self-efcacy and parent Method
support for eating) and dietary intake may differ by gender.
A common shortcoming of adolescent obesity programmes Participants
across Europe is that they often look at adolescents as a homoge-
neous cohort and ignore the growing evidence that boys and girls Ethical approval to conduct the study was obtained from the
differ in their exposure and vulnerability to obesity and eating National University of Galway Ethics Review Board. Criteria for
problems (World Health Organization, 2011). They are affected inclusion were adolescents aged between 13 and 18 years, parental
50 A. Fitzgerald et al. / Appetite 63 (2013) 4858

written consent and informed consent from participants. Criteria that capture the major dietary patterns in the population. For each
for exclusion included participants with multiple or severe food factor, scores are obtained that dene the position of each person
allergies (n = 2). Based on these criteria, 483 participants (264 along a gradient (Reedy et al., 2010). For the current PCA, 98 foods
boys) aged between 13 and 18 years (M = 15.06 years, SD = 1.40) items were reduced to 31 broader food groups by combining items
took part in the study. The sample was predominantly Irish of similar nutrient composition and type. The food groups were
(89%). Based on International Obesity Taskforce criteria (Cole, created based on a review of food group lists in previous Irish re-
Bellizzi, Flegal, & Dietz, 2000), 4.4% of participants were under- search deriving dietary patterns among adults (Hearty & Gibney,
weight, 70.3% were a healthy weight, 21.0% were overweight and 2008).
4.2% were obese. The current study aimed to control for extraneous
sources of variance in the proposed model; therefore, demographic Measures of hypothesized predictors
variables including parent education, age and BMI were included Self-efcacy for healthy eating. Self-efcacy to make healthy food
as covariates in the model testing.1 There were no missing data choices was measured using a nine-item scale developed for use
for age or BMI, however, 5.6% (n = 27) of parental education data with American adolescents (French, Story, Neumark-Sztainer, Fulk-
was missing, therefore, only participants with completed data for erson, & Hannan, 2001). The self-efcacy scale was used to assess
this variable were included in the nal sample (n = 456). Of this participants self-condence for making healthy food choices in so-
sample, those who had more than 10% of missing FFQ data were ex- cial situations (e.g., with friends), in emotional situations (e.g.,
cluded from analyses (n = 11). A nal sample of 445 participants was bored) and in normal situations (e.g., at family dinner). Responses
retained for analyses. were rated on a ve-point scale ranging from not at all sure to
very sure and scores ranged from 9 to 45. Testretest reliability
Procedures for the summary scale score was found to be high (0.83) among
4746 adolescents (French et al., 2001). Adequate-to-high reliability
Participants were recruited from randomly selected post-pri- was also found for the social situation (SS) score (0.66), emotional
mary schools from the Department of Education and Skills pub- situation (ES) score (0.85), and normal situation (NS) score (0.66).
lished list of schools in the West of Ireland. Nineteen post- In the present study, alpha for the summary scale score was 0.73
primary schools were contacted in order to secure the consent of (n = 445). Alpha reliabilities for the three subscales were low
13 schools; (overall school consent rate was 68.42%). Most schools (SS = 0.52, ES = 0.57, NS = 0.56). The reliability problem may be
(97%) were designated as non-disadvantaged. During an initial vis- addressed by employing SEM, where self-efcacy is used as a latent
it, students were provided with a study packet containing an intro- variable, and latent variables are error free (Kline, 2005). The use of
duction letter, an information sheet and a parental consent form. latent variable models means that the relationships in the overall
The individual-level consent rate was 57.92% (ranging from model are not inuenced by measurement error (Bollen, 1989).
43.68% to 77.27% depending on the school). Social support measure for healthy eating. Social support for
healthy eating was measured by an eight-item scale adapted from
Data collection a measure of Diet-specic Social Support for Adolescents (DSSA;
Stanton et al., 2007). The specic items that were included in the
After the researcher obtained written consent from the childs scale are indicated in Table 1. Participants rated on a ve-point Lik-
parent and informed assent from students, self-reported measures ert scale (never to very often) how often a parent or peer per-
were administered during classtime in school. Data collection took formed a behaviour that was supportive of healthy eating. The
place from January to June 2009. parent- and peer-support scales for healthy eating were scored
separately. Total scores for each scale ranged from 4 to 20. Alpha
Measures values for the two scales in the current sample were acceptable:
Basic demographic information. Parents provided demographic 0.73 (n = 445) for parent support for healthy eating and 0.78
information (e.g., childs age and gender, parent education) on (n = 445) for peer support for healthy eating.
the parental consent form. Social support measure for unhealthy eating. A six-item measure
Dietary intake. Dietary intake over the past month was assessed of social support for unhealthy eating was developed based on a
with a 98 item self-administered Food Frequency Questionnaire modication of the DSSA (Stanton et al., 2007). This measure was
(FFQ). The FFQ was selected to collect usual dietary intake as it used to assess how often a parent or peer performed a behaviour
is cost-efcient, easy to administer to a large sample of adoles- that was supportive of unhealthy eating (see Table 1 for specic
cents, has good response compliance, and has shown to be a valid items included in the measure). A denition and examples of com-
and reproducible dietary assessment method among adolescents mon high-fat foods was provided in the instructions. Both the par-
(Rankin, Hanekom, Wright, & MacIntyre, 2010). The FFQ was devel- ent and peer support scale for unhealthy eating were scored
oped for the present study as no previously validated measures separately, and scores ranged from 3 to 15. Alphas for the parent
were available for assessing usual dietary intake in Irish adoles- and peer support scale for unhealthy eating were 0.60 and 0.64,
cents. The FFQ food list was informed by national studies of Irish respectively (n = 445).
childrens and adolescents diets (e.g., National Teens Food Survey, Body mass index. Height and weight were each assessed twice
2008), 24 h food diaries completed in a previous study (Fitzgerald, using a Leicester height rod and a Seca Model 889 medical weight
Heary, Nixon & Kelly, 2010), input from nutrition experts and pilot scale. Anthropometric measurements were made according to
testing of the FFQ. The present study used factor analysis to derive standard methods for body weight and height (Lohman, Roche, &
dietary patterns. Factor analysis (or principal component analysis Martorell, 1998). BMI was calculated as kg/m2. To dene over-
(PCA)) examines the correlation matrix of food variables and weight and obesity, the IOTF sex- and age-specic BMI cut-offs
searches for underlying dimensions (patterns or factors) that ex- were used (Cole et al., 2000).
plain most of the variation in the data (Hu, 2002). Thus, a large
number of food variables are reduced to a smaller set of variables Analytic plan

1
Covariates serve to hold constant the inuences of variables not otherwise
Prior to analysis, data were screened for normality and outliers.
explicitly modelled and they control for extraneous sources of variance in a proposed All survey items assessing hypothesized correlates of dietary
model (Fletcher, Selgrade, & Germano, 2006). intake showed skewness and kurtosis within an acceptable range
 A. Fitzgerald et al. / Appetite 63 (2013) 4858 51

Table 1
Descriptive statistics and factor loadings for model variables.

Latent construct and indicators Final Sample (N = 445) Boys (n = 237) Girls (n = 208) Factor loadings for multi- T value (df)
M (SD) M (SD) M (SD) group Model
Age 15.13 (1.41) 15.02 (1.43) 15.24 (1.38) t(443) = 1.61,
p = .98
Parent education
n (%) n (%) n (%) v2 (6, 445) = .18,
p > .05,
No Schooling 1 (.2) 1 (.4) 0
Primary level or less 17 (3.8) 8 (3.4) 9 (4.3)
Second level 186 (41.8) 90 (38.0) 96 (46.2)
Third level 241 (54.2) 138 (58.2) 103 (49.6)
Parent support for healthy eating
given you ideas on how to eat healthier foods 3.54 (1.09) 3.60 (1.10) 3.48 (1.08) .72 t(443) = 1.14,
p = .80
offered you low-fat snacks 3.27 (1.23) 3.30 (1.21) 3.24 (1.25) .56 t(443) = .54,
p = .71
encouraged you to stay away from high-fat foods 3.37 (1.13) 3.37 (1.12) 3.37 (1.14) .62 t(443) = .06,
p = .80
talked with you about eating more healthy foods 2.86 (1.14) 2.97 (1.16) 2.74 (1.10) .62 t(443) = 2.21,
p < .81
Peer support for healthy eating
given you ideas on how to eat healthier foods 2.01 (1.09) 1.89 (1.08) 2.14 (1.11) .74 t(443) = 2.42,
p = .58
offered you low-fat snacks 2.01 (1.10) 1.87 (1.06) 2.18 (1.14) .57 t(443) = 3.04,
p = .18
encouraged you to stay away from high-fat foods 1.89 (1.03) 1.88 (1.11) 1.89 (.95) .78 t(443) = .21,
p = .001**
talked with you about eating more healthy foods 1.83 (1.03) 1.79 (1.02) 1.88 (1.04) .67 t(443) = .85,
p = .98
Parent support for unhealthy eating
offered you high-fat foods 2.42 (.92) 2.52 (.87) 2.32 (.96) .54 t(443) = 2.32,
p = .30
encouraged you to eat high-fat foods 1.61 (.78) 1.67 (.77) 1.55 (.78) .67 t(443) = 1.59,
p = .79
said nice things about the sweet or high-fat foods 1.89 (.95) 1.85 (.95) 1.95 (.96) .55 t(443) = 1.10,
you were eating p = .90
said nice things about the sweet or high-fat foods 2.94 (1.27) 2.68 (1.21) 2.80 (1.12) .58 t(443) = 1.09,
you were eating p = .08
Peer support for unhealthy eating
offered you high-fat foods 3.30 (1.14) 3.16 (1.09) 3.26 (1.09) .62 t(443) = .70,
p = .98
encouraged you to eat high-fat foods 2.29 (1.20) 2.20 (1.13) 2.28 (1.12) .62 t(443) = 1.10,
p = .08
said nice things about the sweet or high-fat foods 2.94 (1.27) 2.68 (1.21) 2.80 (1.12) .58 t(443) = 1.09,
you were eating p = .08
Self-efcacy (SE) for healthy eating
SE in social situations 7.71 (2.59) 7.28 (2.58) 8.20 (2.51) .63 t(440) = 3.75,
p = .78
SE in emotional situations 8.27 (2.86) 8.52 (2.86) 7.99 (2.84) .59 t(440) = 2.00,
p = .58
SE in normal situations 11.09 (2.57) 10.87 (2.54) 11.34 (2.60) .72 t(438) = 1.95,
p = .44
Unhealthy food intake
Food Group (FG) Confectionary 6.34 (5.99) 5.97 (5.32) 6.76 (6.66) .65 t(443) = 1.41,
p = .02*
FGBiscuits 4.42 (5.02) 4.44 (4.86) 4.39 (5.22) .70 t(443) = .12,
p = .30
FGDesserts 3.51 (4.58) 3.92 (5.60) 3.04 (2.98) .58 t(443) = 2.05,
p = .001**
FGSnacks 3.55 (3.66) 3.74 (3.77) 3.32 (3.54) .57 t(443) = 1.21,
p = 13
Healthy food intake
FGFruit 15.20 (14.54) 14.04 (14.06) 16.52 (14.99) .55 t(443) = 1.80,
p = .57
FGVegetables 19.12 (12.92) 18.07 (12.85) 20.32 (12.92) .62 t(443) = 1.84,
p = .25
FGBrown Bread 3.73 (3.58) 3.61 (3.68) 3.88 (3.48) .43 t(443) = .80,
p = .68
FGLow Energy Drinks 15.39 (7.82) 14.32 (7.47) 16.61 (8.04) .40 t(443) = 3.12,
p = .10
*
Signicance <.01.
**
Signicance <.001.
52 A. Fitzgerald et al. / Appetite 63 (2013) 4858

(Kline, 2005). The FFQ variables exhibited positive skewness and Robinson, Crozier, Inskip, and the SWS Study Group, 2008), the FFQ
high kurtosis, however, food intake variables are not expected to food items were collapsed into a smaller number of food groups for
be normally distributed in the population. If a variable is not ex- entry into the factor analysis. A PCA, using an orthogonal rotation
pected to be normally distributed, an estimation method can be se- (varimax) was conducted on the reported frequencies of consump-
lected that addresses the non-normality (Tabachnick & Fidell, tion of 31 standardized food groups to identify a viable factor
2001). Among the nal sample (n = 445), there were nominal structure based on a randomized split of the data in the sample
amounts of missing item-level data ranging from .2% to 2.6%. Miss- (n = 222). The number of factors to be retained was determined
ing data were imputed using the Expectation Maximization (EM) by a convergence of criteria according to Gorsuch (1983), including
algorithm, a statistical technique for imputing missing data that eigenvalues >1, the scree plot, parallel analysis and theoretical
employs an iterative estimation procedure to converge at a maxi- interpretability of the resulting factor structure. PCA yielded a
mum-likelihood estimate that averages over the distribution of two-factor solution which accounted for 47.99% of the total vari-
missing values (Schafer, 1997). ance (factor 1 = 26.32% (2.11); factor 2 = 21.67% (1.73). The rst
In the hypothesized model, self-efcacy and parent and peer factor termed unhealthy food intake was characterized by high in-
support for healthy and unhealthy eating variables and dietary pat- takes of snacks, sweets, biscuits, and desserts. The second factor
terns were tested as latent constructs in the structural model to termed healthy food intake was characterized by high intakes of
control for measurement error. Demographic covariates (age, fruit, vegetables, brown bread, and low-energy drinks.
BMI, parent education) were tested as directly observed variables. The next step was to conduct a CFA with maximum likelihood
The hypothesized conceptual model was specied and estimated estimation procedures with the remaining sample (n = 223) to
using LISREL 8.8 (Jreskog & Srbom, 2006a). A covariance matrix cross-validate the two-factor structure obtained using PCA. The
and an asymptotic weight matrix were computed using PRELIS two-factor model provided a good t to the data, v2
2.8 (Jreskog & Srbom, 2006b) and the model parameters esti- (df = 19) = 26.08, p = .13, RMSEA = .041 (90% condence interval
mated using robust maximum likelihood (Satorra & Bentler, 2001). (CI): .00.076), IFI = .98, CFI = .98, and SRMR = .048 (see Appendix
Next, mediation analyses were conducted. According to Holm- A).
beck (1997), assuming there is a latent predictor variable (A), a la-
tent mediator variable (B), and a latent outcome variable (C), one Model testing
would rst assess the t of the direct effects model (A ? C). Assum-
ing an adequate t, the t of the overall model is then examined Demographic and study variable means and standard devia-
(A ? B ? C). Assuming this model provides an adequate t, the tions for the overall sample and for boys and girls are presented
A ? B and B ? C path coefcients are then examined. The A ? C, in Table 1. The standardized factor loadings for each of the mea-
A ? B, and B ? C paths should all be signicant in the directions sured variables are presented in Table 1. All factor loadings are po-
predicted. If these paths are signicant, the nal step is to assess sitive and statistically signicant (p < .05). The correlations
the t of the A ? B ? C model under two conditions: (a) when between the variables were mostly signicant but some correla-
the A ? C path is constrained to zero, and (b) when the A ? C path tions were modest. As multiple direct and indirect relationships
is not constrained. were tested, all theoretically meaningful variables were retained
Finally, multi-group analysis was conducted to examine whether for model building, even if they showed modest correlation.
gender moderated the relationship between self-efcacy and parent
and peer support for healthy and unhealthy eating and dietary in- Demographic covariates
take patterns. That is, separate models for boys and girls were spec- Potential demographic covariates were determined by examin-
ied simultaneously within the same overall model. Each model ing associations between age, BMI and parent education and die-
was initially run with all paths xed to be invariant across genders tary intake variables. There were signicant associations between
and tested for gender differences by freeing one path at a time in a each of the covariates and dietary variables, therefore, all covari-
stepwise manner according to the modication indices, and ates were included in subsequent analyses. To control for these
comparing the model with a fully constrained model (see Tildesley variables as possible confounding inuences, paths representing
& Andrews, 2008). If a structural path was signicant (p < .05) for their direct and indirect effects on all latent variables were incor-
either gender, it was included in the nal multigroup model. porated into analyses.
Based on Hoyle and Panters recommendations (1995), the ade-
quacy of model t was assessed using the SatorraBentler scaled Testing assumptions for mediation analysis
chi-square (v2), the Incremental Fit Index (IFI: Bollen, 1989), the In line with Holmbecks (1997) recommendations, the direct ef-
Comparative Fit Index (CFI: Bentler, 1990) and Root Mean Square fects model (A ? C) showed a good t to the data, v2
Error of Approximation (RMSEA; Steiger, 1990). As the v2 statistic (df = 195) = 305.92, p < .001, RMSEA = .034 (90% condence inter-
is sensitive to sample size and violations of normality (Kline, 2005), val (CI): .026.039), CFI = .96, and SRMR = .046 (Model 1). The path
it is reported for completeness, but it is not used to draw specic coefcients from peer support for unhealthy eating to unhealthy
conclusions about the model t (Hu & Bentler, 1999). Well tting food intake (b = .33, p < .01), parent support for unhealthy eating
models should have values that are 60.05 for the RMSEA and val- to unhealthy food intake (b = .21, p < .05), and parent support for
ues that are very close to 0.90 for the IFI and CFI (Hu & Bentler, healthy eating to healthy food intake (b = .15, p < .05) were signif-
1999). A value less than or equal to .05 for the Standardized Root icant. All other path coefcients were not signicant. Next, the t
Mean Square Residual is considered a good t (Kline, 2005). of the overall A ? B ? C model was tested and provided a good
t to the data (v2 (df = 255) = 388.36, p < .001, RMSEA = .033 (90%
condence interval (CI): .026.039), CFI = .96 (Model 2). The path
Results coefcients for this model were examined: peer support for un-
healthy eating was signicantly related to unhealthy food intake
Dietary outcomes (b = .30, p < .01), parent support for unhealthy eating was signi-
cantly related to self-efcacy (b = .32, p < .01) and peer support
A Principal Component Analysis (PCA) was conducted in order for healthy eating was signicantly related to self-efcacy
to generate hypotheses about underlying dietary patterns (b = .14, p < .05). Self efcacy was related to both unhealthy
measured by the FFQ. In line with recommendations (e.g., Borland, (B = .27, p < .01) and healthy (b = .37, p < .01) dietary patterns.
 A. Fitzgerald et al. / Appetite 63 (2013) 4858 53

At this point, the A ? C, A ? B, and B ? C paths should all be signif- on the modication indices for the multiple group model, the
icant in the directions predicted. However, none of the social sup- equality constraint across groups for the path from peer support
port variables were found to be signicantly related to both the for healthy eating to self-efcacy for healthy eating was removed
mediator (self-efcacy) and the outcome variable (either healthy (Model 1.5) and the model provided a good t to the data, v2
or unhealthy dietary pattern) (see Fig. 1). Therefore, there was no (579, N = 445) = 780.05, p < .05, RMSEA = .040 (90% CI: .032; .047),
evidence to support the mediation hypothesis and no further anal- CFI = .94, IFI = .94. The standardized path coefcient between
yses to test the mediation hypothesis were conducted. self-reported peer support for healthy eating and self-efcacy
was signicant for boys (p < .01) but not for girls (p = ns). There
were no other MIs greater than 5. The nal multi-group model
Testing model t by gender with covariates included provided a satisfactory t to the data
Next, multi-group analyses were conducted to statistically eval- (Model 1.6), v2 (714, N = 445) = 1129.80, p < .05, RMSEA = .051
uate the t of the hypothesized model of dietary intake and its (90% CI: .046; .057), CFI = .88, IFI = .88. As the demographic covar-
invariance for boys and girls. Six models were specied and tested. iates were signicantly associated with the dietary outcomes in
The summary statistics for tested models in multigroup analyses initial analyses, the nal model included these covariates so as to
are shown in Table 2. Model 1.1 specied the measurement control for their potential confounding effect on the relationship
model where factor loading, factor variances and covariances and between the latent constructs of social support and self-efcacy
unique variances were allowed to vary across groups, v2 (561, and dietary outcomes. The nal multigroup model t the data
N = 445) = 723.88, p < .05, RMSEA = .036 (90% CI: .028; .044), and no signicant covariances were needed to support the nal
CFI = .95, IFI = .95. However, measurement invariance among model. For simplicity, the non-signicant paths were not shown
groups should be established prior to testing the difference in in the nal multigroup model in Fig. 2.
the structural model across gender. Therefore Model 1.2 included For the nal multi-group model, higher peer support for un-
equality constraints on the factors loading, factor variances and healthy eating (b = .18, p < .05), and lower self-efcacy beliefs were
covariances and unique variances across groups. This model dem- associated with unhealthy intake (b = .29, p < .01). Self-efcacy
onstrated a good t to the data, v2 (579, N = 445) = 772.37, p < .05, was the only latent construct associated with healthy intake
RMSEA = .039 (90% CI: .031; .046), CFI = .94, IFI = .94. Therefore, in (b = .33, p < .01). Parent support for healthy eating was associated
testing for the invariance of the structural model across gender, all with self-efcacy (b = .10, p < .05). With regard to gender effects,
latent variable factor loadings in the measurement model were the path between peer support for healthy eating and self-efcacy
constrained to be equal. was signicant for boys (b = .18, p < .01) but not for girls
Next, multigroup analysis was conducted to test whether the (b = .10). The latent construct of parent support for unhealthy
structural model differed by gender. For simplicity, all covariates eating was not signicantly associated with any of the variables.
were removed from the multigroup analyses of the structural mod- Figure 2 provides details on the signicant associations from the
el until the nal multigroup model. The multiple group model with multi-group analysis.
no equality constraints across groups (Model 1.3) demonstrated a
good t for the data, v2 (566, N = 445) = 772.06, p < .05,
RMSEA = .041 (90% CI: .033; .048), CFI = .94, IFI = .94. With the Discussion
paths specied to be invariant (Model 1.4) the model t statistics
still indicated acceptable t, v2 (580, N = 445) = 787.07 p < .05, The study ndings provide support for testing SCT predictions
RMSEA = .040 (90% CI: .033; .047), CFI = .94, IFI = .94. To determine in relation to adolescents dietary patterns. In line with previous
whether structural paths differed by gender, one path was freed at dietary studies using SCT (Bere & Klepp, 2004; Gallaway, Jago,
a time in the model based on the modication indices (MI > 5), and Baranowski, Baranowski, & Diamond, 2007), higher self-efcacy
the model was compared with the fully constrained model. Based beliefs were directly related to healthy food intake, and lower

Fig. 1. Partial mediation model of the relationship between hypothesized factors and dietary intake patterns for the overall sample (Model 2) p < .05; 
p < .01.
54 A. Fitzgerald et al. / Appetite 63 (2013) 4858

Table 2
Summary statistics for tested models in multiple group analyses.

Model Parameter Satorra Bentler Chi square (df) MI Chi square RMSEA 90% CI CFI IFI
***
1.1 723.88 (561) 859.87 .036 .028; .044 .95 .95
1.2 772.37 (579) *** 955.86 .039 .031; .046 .94 .94
1.3 772.06 (566) *** 951.50 .041 .033; .048 .94 .94
1.4 787.07 (580) *** 968.37 .040 .033; .047 .94 .94
1.5 Peer support for healthy eating ? self-efcacy 780.05 (579) *** 5.91 961.96 .040 .032; .047 .94 .94
1.6 1129.80 (714) *** 1370.22 .051 .046; .057 .88 .88

MI, Modication Index.

1.1 All factor loadings in the measurement model were estimately freely for boys and girls.
1.2 All factor loadings in the measurement model were constrained to be equal for boys and girls.
1.3 All path coefcients in the structural model were estimated freely.
1.4 All path coefcients in the structural model were constrained to be equal.
1.5 Path coefcient from peer support for healthy eating to self-efcacy to make healthy food choices was estimated freely.
1.6 Final multi-group model with one parameter estimated freely across gender.
***
p < .001.

Fig. 2. Final multi-group model for boys and girls. Standardized parameter estimates in the nal SEM. The estimates for the girls were shown in the parentheses. For clarity,
non-signicant paths are not shown in the above diagram. All paths shown are statistically signicant p < .05; p < .01.

self-efcacy beliefs were directly related to unhealthy food intake. lescents whose peers support them to eat unhealthily reported a
This study tested the conditions necessary for a mediating effect of lower level of self-efcacy, which in turn was associated with a
self-efcacy in the relationship between social support variables higher level of unhealthy eating. This nding suggests that it is
and dietary intake. There was no evidence that self-efcacy medi- important for future research to examine the effect of peer support
ated the relationship between social support variables and dietary for unhealthy eating and dietary intake.
patterns. To date, previous research has provided inconsistent evi- The lack of a mediation effect could be attributed to measure-
dence regarding the mediating role of self-efcacy between social ment of self-efcacy and social support, where the behaviours that
support variables and dietary intake. Some research has reported a the self-efcacy measure was based on were distinct from the
mediating effect for self-efcacy in the relationship between per- behaviours that were utilized in the social support measure. Fur-
ceived parent support and fruit and vegetable intake (Young ther research is needed to clarify these ndings using measures
et al., 2004), while other research has reported mixed effects, of self-efcacy and social support for eating that tap into the same
where self-efcacy mediated the relationship between parent sup- target dietary behaviours. Additionally, the use of more behaviour-
port and calcium intake but did not mediate the relationship be- specic measures of predictor variables that relate to dietary out-
tween peer support and intake (Ievers-Landis et al., 2003). In comes of interest in model testing are warranted and would likely
another study, Reynolds et al. (1999) reported an indirect effect, explain larger percentages of dietary intake (Neumark-Sztainer,
but no mediated effect, between social variables and intake via Wall, Perry, & Story, 2003).
self-efcacy. In the current study, the relationship between parent Another important aim of this research was to examine the role
support for unhealthy eating and peer support for healthy eating of gender as a moderator of the associations between hypothesized
were directly related to self-efcacy, and self-efcacy in turn was factors and dietary patterns. The multigroup model was similar
related to the dietary outcomes. These ndings suggest that ado- for boys and girls, with one path found to differ across gender.
 A. Fitzgerald et al. / Appetite 63 (2013) 4858 55

Specically, the path between peer support for healthy eating and Fisher & Birch, 2000). However, these studies have mainly focused
self-efcacy was signicantly stronger for boys than girls. on parental controlling and monitoring styles among children. Fur-
The nding of a signicant relationship between peer support ther research would be warranted to investigate the role that par-
for healthy eating and self-efcacy for boys but not girls may sug- ents play in supporting their adolescents to eat unhealthy foods
gest that boys perceptions of peer support for healthy eating is perhaps using multi-informant design and employing a combina-
more closely linked to their condence/ability to eat healthily than tion of self-report and other-report measures (e.g., parents, peers).
it is for girls. It may be that boys whose peer group encourage and Alternatively, capturing food intake objectively through the use of
support one another to consume healthy foods have greater con- weighed food methods would be valuable to explore adolescents
dence to make healthy foods choices, particularly in social settings, dietary intake combined with parental reports of food strategies
such as when eating out with friends or in peer groups. As this employed within the home.
study was exploratory in nature, it is unclear why this gender dif- Contrary to previous research, the current study found no gen-
ference was found. Observational studies on the peer eating envi- der difference in the relationship between parent support for eat-
ronment in schools would be worthwhile to tease out whether ing and dietary intake. Previous studies reporting an association
gender variations in peers provision of support for eating in real- between parent support for healthy eating and boys dietary in-
life settings exist. take, but not girls, focused on specic dietary outcomes such as
The current study extended research by showing that peer sup- fast-food intake (Bauer et al., 2009) or calcium-intake (Larson
port for unhealthy eating was associated with an unhealthy dietary et al., 2006), whereas the current study focused on the intake of
pattern. This nding is supported by qualitative research where combinations of foods. It may be that parents encourage their
adolescents related eating unhealthy/junk foods predominantly son or daughter to eat specic foods depending on their son or
with their friends (Watt & Sheiham, 1997) and peer-related social daughters eating patterns. For example, it may be that parents
pressures were cited as a barrier to healthy eating (Croll et al., encourage their sons, but not their daughters, to consume lower in-
2001). Experimental research also indicates that social facilitation takes of fast food and higher intakes of calcium if they are aware
of eating occurs when children are familiar with each other (Salvy, that their son consumes high intakes of unhealthy foods and low
Vartanian, Coelho, Jarrin, & Pliner, 2008). There is also evidence intakes of healthy foods. It is important for future research to fur-
that the larger peer crowd with which adolescents afliate (e.g., ther examine whether the relationship between parent support for
Jocks, Populars) is related to their eating and weight control behav- eating and dietary intake differ by gender.
iours (MacKey & La Greca, 2007). These ndings suggest that Strengths of this study include the use of food factor scores (die-
involving adolescents close friends and peers in dietary interven- tary patterns) as dependent variables in model testing, the concep-
tions might help to offset the potentially competing inuences of tualization of self-reported social support for eating and a primary
the peer group. focus on the documented gender differences in factors associated
There are several potential explanations for the association be- with dietary patterns. A main weakness of this study was the use
tween peer support for unhealthy eating and dietary intake. While of self-report measures for the variables under investigation. Most
SCT emphasizes that individuals learn from other people, or mod- of the data were from adolescent self-report measures as opposed
els, through observational learning in their social environment, to objective measures and thereby subject to possible attention,
other theories can be drawn upon to understand the role of peers comprehension, memory, and recording errors. Adolescents were
on health behaviours. A number of theories have specied pro- asked to report how frequently their parents and peers performed
cesses by which peers inuence each others behaviour (Brechwald certain supportive behaviours. Adolescents may however be vul-
& Prinstein, 2011). These include selection processes where chil- nerable to all kinds of biases, for example, they may simply not no-
dren may choose friends that are similar in behaviour and beliefs. tice or remember how frequently they or others provide support
Identity based theories suggest that during adolescence in particu- for eating. Adolescents perceptions of others supportive behav-
lar, young people internalize and emulate idealized values and iours may also interact with their own desire for unhealthy foods
behaviours of peers and adhere to social norms established within at a certain moment, or with their actual behaviour. Adolescents
their peer group (Brechwald & Prinstein, 2011). The extent to may not accurately report their parents and peers behaviours be-
which these processes operate in relation to adolescents food con- cause of a failure to recall important events, selective bias, or in
sumption remains relatively underexplored. some cases, intentional distortions (Mash & Wolfe, 2008). Some
There were no direct associations found between parent sup- adolescents may try to make their parents/peers look better or
port for healthy eating and dietary intake variables, however, par- worse than they are. These cognitive biases might be a problem
ent support was associated with self-efcacy, which in turn was particularly during puberty as adolescents begin to develop auton-
related to an unhealthy food intake. This lack of a signicant direct omy from parents and the role of peers becomes increasing impor-
association may be attributed to adolescents increasing autonomy tant and should be considered as a limitation of the current
in making food choice decisions away from the home (Story, Neu- research. Future research could use a multi-informant design as
mark-Sztainer, & French, 2002). Longitudinal research has shown data from parents or peers would serve as useful adjuncts to the
that parental encouragement for healthy eating decreased from self-report methodology. The use of self-report measures of both
early to mid adolescence (Bauer, Laska, Fulkerson, & Neumark- hypothesized factors and dietary patterns is another limitation in
Sztainer, 2011). There was also no association found between per- that shared method variance could be inating the strength of
ceived parent support for unhealthy eating and unhealthy dietary these associations. Longitudinal research is needed to determine
pattern. In the current study, adolescents were asked to report whether relationships between hypothesized variables and dietary
their parents support for eating, therefore, results may reect patterns are causal. Such studies are particularly important given
changes in adolescents perceptions rather than changes in paren- our focus on adolescents and their increasing autonomy and
tal behaviour. As adolescents become older they may spend less changing social context.
time with their parents, they may perceive less parent support The present study used a FFQ to assess habitual dietary intake.
for eating than at previous points, regardless of actual changes in Measurement errors inherent to the use of a FFQ for dietary assess-
parental behaviour. Despite the lack of an association between ment are well documented (Livingstone & Robson, 2000). Despite
parental support and dietary intake in this study, there is some evi- this, previous research has observed similar estimates for energy
dence that parents can promote and model unhealthy eating habits intake with multiple 24-h recalls and a FFQ assessing diet over
to their children (Arredondo et al., 2006; Brown & Ogden, 2004; the past year in adolescents (Rockett & Colditz, 1997). The use of
56 A. Fitzgerald et al. / Appetite 63 (2013) 4858

a non-validated FFQ is another limitation, however, the reading To conclude, these ndings offer evidence to suggest that
and face validity of the FFQ was established in a pilot study (Fitz- self-efcacy for healthy eating is associated with a healthy dietary
gerald, Heary, Nixon & Kelly, 2010) and the dimensionality of the pattern, while perceived peer support for unhealthy eating is
FFQ was explored. The cultural sensitivity of the self-efcacy mea- associated with an unhealthy dietary pattern. Recognizing the
sure should also be considered given that the reliabilities of the importance of peer support for unhealthy eating during adoles-
subscales for the current sample were lower compared to an Amer- cence may be important to consider in dietary interventions.
ican sample. Low reliability may also raise questions about the Although the multigroup model was generally similar for boys
validity of the scale and it would be worthwhile to validate the and girls, further research is warranted to draw rm conclusions
self-efcacy scale with an Irish population of adolescents. Finally, about the role of gender as a moderator of the associations.
the present study included a sample of Irish adolescents which
may limit generalizability of ndings to other cultural contexts.
Different social, cultural, economic, policy and regulatory food References
environments might mean that results are not always transferable
Arredondo, E. M., Elder, J. P., Ayala, G. X., Campbell, N., Baquero, B., & Duerksen, S.
from one country to the next. (2006). Is parenting style related to childrens healthy eating and physical
Age was included as a covariate in the model, however it would activity in Latino families? Health Education Research, 21, 862871.
be worthwhile to look at whether age moderates the relationship Ayala, G. X., Baquero, B., Arredondo, E. M., Campbell, N., Larios, S., & Elder, J. P.
(2007). Association between family variables and Mexican American
between hypothesized factors and adolescents dietary patterns. childrens dietary behaviours. Journal of Nutrition Education & Behaviour, 39,
Additionally, given that changes in food habits between early ado- 6269.
lescence and adulthood have been shown (Lake, Mathers, Rugg- Bandura, A. (1986). Social foundations of thought and action. A social cognitive theory.
Englewood Cliffs, NJ: Prentice Hall.
Gunn, & Adamson, 2006), it would be worthwhile to examine gen- Bauer, K. W., Larson, N. I., Nelson, M. C., Story, M., & Neumark-Sztainer, D. (2009).
der as a moderator of the relationship between correlates and die- Socio-environmental, personal and behavioural predictors of fast-food intake
tary intake for specc age groups of adolescents including younger among adolescents. Public Health Nutrition, 12, 17671774.
Bauer, K., Laska, M., Fulkerson, J. A., & Neumark-Sztainer (2011). Longitudinal and
(1315 years) and older (1618 years) adolescents. There was no secular trends in parental encouragement for healthy eating, physical activity,
relationship found between BMI and dietary patterns which may and dieting throughout the adolescent years. Journal of Adolescent Health, 49,
be explained by the cross-sectional nature of the study. Alterna- 306311.
Bentler, P. M. (1990). Comparative t indexes in structural models. Psychological
tively, it may be that BMI does not inuence dietary patterns, or Bulletin, 107, 238246.
that if it does, a longer term prospective design is needed to inves- Bere, E., & Klepp, K. I. (2004). Correlates of fruit and vegetable intake among
tigate this. Norwegian schoolchildren. Parental and self-reports. Public Health Nutrition, 7,
991998.
In line with previous research (Bere & Klepp, 2004; Neumark-
Bollen, K. A. (1989). Structural equations with latent variables. New York: Wiley.
Sztainer et al., 2003), the model accounted for less than a third of Borland, S. E., Robinson, S. M., Crozier, S. R., Inskip, H. M., & the SWS Study Group
the variance in dietary intake. The next step to theoretically ad- (2008). Stability of dietary patterns in young women over a 2-year period.
vance the literature may be to incorporate other SCT constructs European Journal of Clinical Nutrition, 62, 119126.
Brechwald, W. A., & Prinstein, M. J. (2011). Beyond homophily. A decade of advances
that target the physical environment such as school food availabil- in understanding peer inuence processes. Journal of Research on Adolescence,
ity or the neighbourhood food environment into models to explain 21, 166179.
more of the variance in dietary patterns. In one of the rst studies British Medical Association (2003). Adolescent health. London: British Medical
Association.
examining aspects of both the social (peer friendship groups) and Brown, R., & Ogden, J. (2004). Childrens eating attitudes and behaviour: A study of
physical environment (school availability), both peer group snack the modelling and control theories of parental inuence. Health Education
and soft drink consumption were associated with intake particu- Research: Theory & Practice, 19, 261271.
Cerin, E., Barnett, A., & Baranowski, T. (2009). Testing theories of dietary behaviour
larly when availability in the school canteen and vending machine change in youth using the mediating variable model with intervention
was high (Wouters, Larsen, Kremers, Dagnelie, & Geenen, 2010). programs. Journal of Nutrition Education and Behaviour, 41, 309318.
The studys ndings suggest that interventions that address Cheng, S., & Chan, A. C. M. (2004). The multidimensional scale of perceived social
support. Dimensionality and age and gender differences in adolescents.
self-efcacy for making healthy food choices in different situations Personality and Individual Differences, 37, 13591369.
may be benecial for adolescents. Dietary strategies could help Colarossi, L. G. (2001). Adolescent gender differences in social support. Structure,
students to make healthy food choices more readily in social situ- function, and provider type. Social Work Research, 25, 233241.
Cole, T. J., Bellizzi, M. C., Flegal, K. M., & Dietz, W. H. (2000). Establishing a standard
ations with peers, when feeling bored/stressed or when at home.
denition for child overweight and obesity worldwide. International survey.
Adolescents could also be taught skills to develop their personal British Medical Journal, 320, 12401243.
efcacy to resist peer encouragement to consume unhealthy foods. Croll, J. K., Neumark-Sztainer, D., & Story, M. (2001). Healthy eating. What does it
Within the school context, adolescents could be taught self-advo- mean to adolescents? Journal of Nutrition Education, 33, 193198.
Cutler, G. J., Flood, A., Hannan, P. J., Slavin, J. L., & Neumark-Sztainer, D. (2011).
cacy skills including assertiveness and effective communication Association between major patterns of dietary intake and weight status in
skills through role play scenarios with their peers around food. adolescents. British Journal of Nutrition, 13, 18.
These ndings support the recommendation of Wouters et al. Engeset, D., Alsaker, E., Ciampi, A., & Lund, E. (2005). Dietary patterns and lifestyle
factors in the Norwegian EPIC cohort. The Norwegian Women and Cancer
(2010), who argue for the need for interventions to focus on groups (NOWAC) study. European Journal of Clinical Nutrition, 59, 675684.
of friends as well as on the individual. Eschenbeck, H., Kohlmann, C. W., & Lohaus, A. (2007). Gender differences in coping
In addition, in order to create healthy norms amongst adoles- strategies in children and adolescents. Journal of Individual Differences, 28,
1826.
cent peer groups, the peer context needs to be facilitated and sup- Finnerty, T., Reeves, S., Dabinett, J., Jeanes, Y. M., & Vgele, C. (2009). Effects of peer
ported by the broader school, family and community context. inuence on dietary intake and physical activity in schoolchildren. Public Health
Eating is frequently a social occasion. Therefore, the availability Nutrition, 1, 18.
Fisher, J. O., & Birch, L. L. (2000). Parents restrictive feeding practices are associated
of healthy foods in schools and communities where adolescents with young girls negative self-evaluation of eating. Journal of the American
gather socially can enable young people to make healthy food Dietetic Association, 100, 13411346.
choices. At a broader population level, social norm interventions Fitzgerald, A., Heary, C., Nixon, E., & Kelly, C. (2010). Factors inuencing the food
choices of Irish children and adolescents: a qualitative investigation. Health
could also be explored in an effort to change adolescents percep-
Promotion International, 25, 289298.
tions of the beliefs and behaviours of other young people regarding Fletcher, T. D., Selgrade, K. A., & Germano, L. M. (2006, May). On the use of partial
food, particularly junk food. A recent social norm marketing cam- covariances in structural modeling. Poster session presented at the annual
paign had signicant effects on normative misperceptions and conference of the Society for Industrial and Organizational Psychology, Dallas,
TX.
drinking behaviour (Perkins, Linkenbach, Lewis, & Neighbors, French, S. A., Story, M., Neumark-Sztainer, D., Fulkerson, J. A., & Hannan, P. (2001).
2011). Fast food restaurant use among adolescents. Associations with nutrient intake,
 A. Fitzgerald et al. / Appetite 63 (2013) 4858 57

food choices and behavioural and psychosocial variables. International Journal of Lubans, D. R., Plotnikoff, R. C., Morgan, P. J., Dewar, D., Costigan, S., & Collins, C. E.
Obesity, 25, 18231833. (2012). Explaining dietary intake in adolescent girls from disadvantaged
Frenn, M., Malin, S., & Bansal, N. K. (2003). Stage-based interventions for low-fat secondary schools. A test of Social Cognitive Theory. Appetite, 58, 517524.
diet with middle school students. Journal of Pediatric Nursing, 18, 3645. Mash, E. J., & Wolfe, D. A. (2008). Abnormal child psychology (4th ed.). Belmont, CA:
Gallaway, M. S., Jago, R., Baranowski, T., Baranowski, J. C., & Diamond, P. M. (2007). Wadsworth.
Psychosocial and demographic predictors of fruit, juice and vegetable MacKey, E. R., & La Greca, A. M. (2007). Adolescents eating, exercise and weight
consumption among 1114-year-old Boy Scouts. Public Health Nutrition, 10, control behaviors: Does peer crowd afliation play a role? Journal of Pediatric
15081514. Psychology, 32, 1323.
Gentile, D., Welk, G., Eisenmann, J., Reimer, R., Walsh, D., Russell, D., et al. (2009). McClain, A. D., Chappuis, C., Nguyen-Rodriguez, S. T., Yaroch, A. L., & Spruijt-Metz, D.
Evaluation of a multiple ecological level child obesity prevention program. (2009). Psychosocial correlates of eating behaviour in children and adolescents.
Switch what you Do, View, and Chew. BMC Medicine, 7, 4960. A review. International Journal of Behavioural Nutrition and Physical Activity, 6,
Gorsuch, R. L. (1983). Factor analysis (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum 5474.
Associates. Mikkila, V., Rasanen, L., Raitakari, O. T., Pietnen, P., & Vikari, J. (2005).
Gould, M., Jasik, C., Lustig, R., & Garber, A. (2009). A clinic-based nutrition Consistent dietary patterns identied from childhood to adulthood. The
intervention improves self-efcacy to change behaviour in obese adolescents Cardiovascular Risk in Young Finns Study. British Journal of Nutrition, 58,
and parents. Journal of Adolescent Health, 44, S39S40. 10381045.
Groth, M. V., Fagt, S., Stockmarr, A., Matthiessen, J., & Biltoft-Jensen, A. (2009). Neumark-Sztainer, D., Story, M., Hannan, P. J., & Croll, J. (2002). Overweight status
Dimensions of socioeconomic position related to body mass index and obesity and eating patterns among adolescents. Where do youths stand in comparison
among Danish women and men. Scandinavian Journal of Public Health, 37, with the healthy people 2010 objectives? American Journal of Public Health, 92,
418426. 844851.
Hanson, N. I., Neumark-Sztainer, D., Eisenberg, M. E., Story, M., & Wall, M. (2005). Neumark-Sztainer, D., Wall, M., Perry, C., & Story, M. (2003). Correlates of fruit and
Associations between parental report of the home food environment and vegetable intake among adolescents. Findings from Project EAT. Preventive
adolescent intakes of fruits, vegetables and dairy foods. Public Health Nutrition, Medicine, 37, 198208.
8, 7785. Perkins, H. W., Linkenbach, J. W., Lewis, L. A., & Neighbors, C. (2011). Effectiveness of
Hearty, A. P., & Gibney, M. J. (2008). Comparison of cluster and principal component social norms media marketing in reducing drinking and driving. A statewide
analysis techniques to derive dietary patterns in Irish adults. British Journal of campaign. Addictive Behaviours, 35, 866874.
Nutrition, 101, 598608. Rankin, D., Hanekom, S. M., Wright, H. H., & MacIntyre, U. E. (2010). Dietary
Hedley, A. A., Ogden, C. L., Johnson, C. L., Carroll, M. D., Curtin, L. R., & Flegal, K. M. assessment methodology for adolescents. A review of reproducibility and
(2004). Prevalence of overweight and obesity among US children, adolescents, validation studies. South African Journal of Clinical Nutrition, 23, 6574.
and adults, 19992002. Journal of the American Medical Association, 291, Reedy, J., Wirfalt, E., Flood, A., Mitrou, P. N., Krebs-Smith, S. M., Kipnis, V., et al.
28472850. (2010). Comparing 3 dietary pattern methods- cluster analysis, factor analysis,
Holmbeck, G. N. (1997). Toward terminological, conceptual, and statistical clarity in and index analysis- with colorectal cancer risk. American Journal of
the study of mediators and moderators. Examples from the child-clinical and Epidemiology, 15, 479487.
pediatric psychology literatures. Journal of Consulting and Clinical Psychology, 65, Reynolds, K., Hinton, A., Shewchuck, R., & Hickey, C. (1999). Social cognitive model
599610. of fruit and vegetable consumption in elementary school children. Journal of
Hoyle, R. H., & Panter, A. T. (1995). Writing about structural equation models. In R. Nutrition Education & Behavior, 31, 2330.
H. Hoyle (Ed.), Structural equation modeling. Concepts, issues and applications Rinderknecht, K., & Smith, C. (2004). Social cognitive theory in an after-school
(pp. 158198). London: Sage. nutrition intervention for urban Native American youth. Journal of Nutrition
Hu, F. B. (2002). Dietary pattern analysis. A new direction in nutritional Education & Behaviour, 36, 298304.
epidemiology. Current Opinion in Lipidology, 13, 39. Rockett, H. R., & Colditz, G. A. (1997). Assessing diets of children and adolescents.
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for t indexes in covariance structure American Journal of Clinical Nutrition, V65, 1116S1122S.
analysis. Conventional criteria versus new alternatives. Structural Equation Salvy, S. J., Vartanian, L. R., Coelho, J. S., Jarrin, D., & Pliner, P. P. (2008). The role of
Modeling, 6, 155. familiarity on modeling of eating and food consumption in children. Appetite,
Ievers-Landis, C. E., Burant, C., Drotar, D., Morgan, L., Trapl, E. S., & Kwoh, C. K. 50, 514518.
(2003). Social support, knowledge, and self-efcacy as correlates of Sanchez-Villegas, A., Delgado-Rodriguez, M., Martinez-Gonzalez, M. A., Irala-
osteoporosis preventive behaviours among preadolescent females. Journal of Estevez & J. for the SUN group (2003). Gender, age, socio-demographic and
Pediatric Psychology, 28, 335345. lifestyle factors associated with major dietary patterns in the Spanish Project
Institute of Medicine (2007). Nutrition Standards for food in schools. Leading the way SUN. European Journal of Clinical Nutrition, 57, 285292.
towards healthier youth. Washington, DC: National Academies Press. Satorra, A., & Bentler, P. M. (2001). A scaled difference chi-square test statistic for
Irish Universities Nutrition Alliance (2008). National Teens Food Survey. Main report. moment structure analysis. Psychometrika, 66, 507514.
<http://www.iuna.net/index.php?option=com_content&view=article&id=8& Schafer, J. L. (1997). Analysis of incomplete multivariate data. New York: Chapman &
Itemid=20> Retrieved 19.12.09. Hall.
Jreskog, K. G., & Srbom, D., (2006a). LISREL 8 (version 8.8). Chicago, IL: Scientic Stanton, C. A., Green, S. L., & Fries, E. A. (2007). Diet-specic social support
Software. among rural adolescents. Journal of Nutrition Education & Behaviour, 39,
Jreskog, K. G., & Srbom, D., (2006b). PRELIS 2 (version 2.8). Chicago, IL: Scientic 214218.
Software. Steiger, J. H. (1990). Structural model evaluation and modication. An interval
Kline, R. B. (2005). Principles and practice of structural equation modeling (2nd ed.). estimation approach. Multivariate Behavioural Research, 25, 173180.
London: Guildford Press. Story, M., Neumark-Sztainer, D., & French, S. (2002). Individual and environmental
Lake, A. A., Mathers, J. C., Rugg-Gunn, A. J., & Adamson, A. J. (2006). Longitudinal inuences on adolescent eating behaviours. Journal of the American Dietetic
change in food habits between adolescence (1112 years) and adulthood (32 Association, 102, S40S51.
33 years). The ASH30 Study. Journal of Public Health, 28, 1016. Tabachnick, B. G., & Fidell, L. S. (2001). Understanding multivariate statistics (3rd ed.).
Larson, N. I., Neumark-Sztainer, D., Harnack, L. J., Wall, M. M., Story, M. T., & New York: Harper Collins.
Eisenberg, M. E. (2008). Fruit and vegetable intake correlates during the Tildesley, E. A., & Andrews, J. A. (2008). The development of childrens intentions to
transition to young adulthood. American Journal of Preventive Medicine, 35, use alcohol. Direct and indirect effects of parent alcohol use and parenting
3337. behaviours. Psychology of Addictive Behaviours, 22, 326339.
Larson, N. I., Neumark-Sztainer, D., Harnack, L. J., Wall, M. M., Story, M. T., & Tucker, K. L. (2010). Dietary patterns, approaches, and multicultural perspective.
Eisenberg, M. E. (2009). Fruit and vegetable intake correlates during the Applied Physiology, Nutrition and Metabolism, 35, 211218.
transition to young adulthood. American Journal of Preventive Medicine, 35, Watt, R. G., & Sheiham, A. (1997). Towards an understanding of young peoples
3337. conceptualisation of food and eating. Health Education, 56, 340349.
Larson, N. I., Story, M., Wall, M., & Neumark-Sztainer, D. (2006). Calcium and dairy Wind, M., de Bourdeaudhuij, I., te Velde, S. J., Sandvik, C., Due, P., Klepp, K. I., et al.
intakes of adolescents are associated with their home environment, taste (2006). Correlates of fruit and vegetable consumption among 11-year-old
preferences, personal health beliefs, and meal patterns. Journal of the American Belgian-Flemish and Dutch schoolchildren. Journal of Nutrition Education and
Dietetic Association, 106, 18161824. Behaviour, 38, 211221.
Lien, N., Jacobs, D. R., & Klepp, K. I. (2002). Exploring predictors of eating behaviour World Health Organization (2011). Evidence for gender responsive actions to prevent
among adolescents by gender and socio-economic status. Public Health and manage overweight and obesity. Young peoples health as a whole-of-society
Nutrition, 5, 671681. response. WHO Regional Ofce for Europe.
Lien, N., Lytle, L. A., & Klepp, K. I. (2001). Stability in consumption of fruit, Wouters, E. J., Larsen, J. K., Kremers, S. P., Dagnelie, P. C., & Geenen, R. (2010). Peer
vegetables, and sugary foods in a cohort from age 14 to age 21. Preventive inuence on snacking behaviour in adolescence. Appetite, 55, 1117.
Medicine, 33, 217226. Young, E. M., Fors, S. W., & Hayes, D. M. (2004). Associations between perceived
Livingstone, M. B. E., & Robson, P. J. (2000). Measurement of dietary intake in parent behaviors and middle school student fruit and vegetable consumption.
children. Proceedings of the Nutrition Society, 59, 279293. Journal of Nutrition Education & Behavior, 36, 28.
Lobstein, T. J., James, W. P. T., & Cole, T. J. (2003). Increasing levels of excess weight Zabinski, M. F., Daly, T., Norman, G. J., Rupp, J. W., Calfas, K. J., Sallis, J. F., et al.
among children in England. International Journal of Obesity, 27, 11361138. (2006). Psychosocial correlates of fruit, vegetable, and dietary fat intake among
Lohman, T. G., Roche, A. F., & Martorell, R. (1998). Anthropometric standardization adolescent boys and girls. Journal of the American Dietetic Association, 106,
reference manual. Champaign, IL: Human Kinetics. 814821.
58 A. Fitzgerald et al. / Appetite 63 (2013) 4858

Appendix A

See Fig. A.

Fig. A. Conrmatory factor analysis model of the two-factor solution derived from PCA (n = 223).