Sei sulla pagina 1di 13

Psychology of Popular Media Culture 2012 American Psychological Association

2013, Vol. 2, No. 1, 38 50 2160-4134/13/$12.00 DOI: 10.1037/a0030277

Development and Validation of a Social Media Use


Integration Scale

Michael A. Jenkins-Guarnieri, Stephen L. Wright, and Brian Johnson


University of Northern Colorado

The present study developed a scale of online social media use that measures the
integration of the social behavior and daily routines of users, along with the importance
of and emotional connection to this use. Using a sample of 616 emerging adults in
college, exploratory factor analysis was conducted with a calibration sample of 308
participants and confirmatory factor analysis was conducted using an equal hold-out
sample to yield a final 10-item two-factor Social Media Use Integration Scale (SMUIS).
Strong reliability evidence was found for data collected with the total scale ( .914),
the first 6-item subscale called Social Integration and Emotional Connection (SIEC)
( .893), and the second 4-item subscale entitled Integration into Social Routines
(ISR) ( .828). Testretest over a 3-week period suggested that SMUIS responses
remained stable, with reliability correlations of r .803 for total scale, r .804 for
subscale 1, and r .676 for subscale 2. In addition, high correlations with previously
published social media use measures provided convergent validity evidence, whereas
nonsignificant correlations between the SMUIS subscales and other measures unrelated
to online social media use offered discriminant validity evidence. The SMUIS was first
developed to measure Facebook use; however, it was purposefully designed to be
adapted to measure other forms of online social media use. Implications for future
research and practice are discussed.

Keywords: Facebook, online social media, college students, measurement scale

Internet use for communication and social quickly become the most popular SNS for
behavior is becoming increasingly integrated young adults, especially those in college
into the lives of North Americans (Correa, Hin- (Cheung, Chiu, & Lee, 2011). As this online
sley, & de Ziga, 2010; Ross et al., 2009), with social medium is increasingly integrated into
95% of young adults aged 18 to 33 years re- the daily lives and social behavior of young
porting activity online in recent years (Zickurh, adults (Correa et al., 2010; Steinfield, Ellison, &
2010). One specific area that has seen a signif- Lampe, 2008), new research must seek to assess
icant and rapid rise to prominence (Pempek, and understand the nature of using this medium
Yermolayeva, & Calvert, 2009) has been social and its potential implications, especially for so-
networking sites (SNS) such as Facebook.com, cial behavior and development (Brown, 2006).
online social media through which to share ex- It is imperative to use well-developed measures
periences and communicate within social rela- to effectively investigate online social behavior;
tionships (Ross et al., 2009). Facebook has however, recent research on online social media
has often used inadequate measures for opera-
tionalizing social media use and its integration
into social behavior. Furthermore, based on a
This article was published Online First October 22, 2012. thorough literature review, to date, no published
Michael A. Jenkins-Guarnieri, Stephen L. Wright, and scale has been developed following more for-
Brian Johnson, Department of Counseling Psychology, Uni-
versity of Northern Colorado. mal and rigorous methods of scale development
Correspondence concerning this article should be ad- and validation for measuring how social media
dressed to Michael A. Jenkins-Guarnieri, Department of is used and integrated into the daily lives of
Counseling Psychology, University of Northern Colorado,
501 20th Street, McKee Hall 248, Campus Box 131, Gree-
users. Thus, the present study used methodolog-
ley, CO 80639. E-mail: Michael.JenkinsGuarnieri@gmail ically rigorous techniques for scale develop-
.com ment (DeVellis, 2003) to create a novel measure
38
SOCIAL MEDIA USE INTEGRATION SCALE 39

for social media use integration, and established newer, most research on social media use thus
initial validity evidence for its use with emerg- far has used psychometrically weak measures.
ing adult populations. Although intended to be Based on scale development theory (DeVellis,
flexible enough to be adapted for use with mul- 2003), even the most recent research published
tiple manifestations of online social media, this in peer-reviewed journals used somewhat lack-
study focused on Facebook, given its promi- ing assessment measures to operationalize Fa-
nence in North America, especially with col- cebook use. For example, researchers often use
lege-aged adults. a single item assessing factual information
about use, such as estimates of the average daily
Emerging Adults and Social Media Use activity of use in minutes or account logins per
week (e.g., Baker & Oswald, 2010; Litt &
Boyd and Ellison (2007) highlighted the di- Stock, 2011; Ong et al., 2011; Wilson, For-
versity and significant numbers of online social nasier, & White, 2010). Additionally, the for-
media introduced since their emerging forms as mat and response scales for this type of question
early as 1997, as these online communication varied considerably across studies. Single items
tools began to evolve with different features and such as these often perform poorly in measuring
foci. In reviewing the development of SNS, they complex constructs, with poor reliability esti-
highlight MySpace as the first major main- mates and high measurement error. In addition,
stream service, with teenagers joining in record much previous research in this area focused on
numbers and later migrating to the incredibly the behavioral frequency of social media use,
popular Facebook. In conjunction around this yet the nature of these social media services
time, other sites like Twitter, YouTube, and most likely involves a users emotional connec-
LinkedIn emerged and began attracting users to tion to and integration of use in their daily social
consume and share media. These users soon lives (Ellison et al., 2007). Therefore, we de-
reached college age and developed into an fined online social media use as the degree to
emerging adult population, maintaining this which social media is integrated into the social
high level of social activity through social me- behavior and daily routines of users, and the
dia to such an extent that developmental tasks importance of and emotional connection to this
associated with this stage (Arnett, 2000) may be use. This focus represents a unique contribution
enacted in part online. Key developmental ac- of our study to the literature in this area, as our
tivities such as exploring individual identity measure was developed to capture this broader
(Grasmuck, Martin, & Zhao, 2009), social iden- concept of engaged use.
tity (Manago, Graham, Greenfield, & Salim- Some researchers have developed novel mea-
khan, 2008), and sense of self (Peluchette & surement scales for their studies involving so-
Karl, 2010) appear to be carried out through cial media, but did not conduct rigorous psy-
social media and SNS such as Facebook. In chometric analyses before using data collected
conjunction, emerging adulthood populations from their new measures to answer subsequent
often initiate and maintain meaningful relation- research questions (e.g., Carpenter, Green, &
ships through Facebook, for example, commu- LaFlam, 2011; Kwon & Wen, 2010; Ong et al.,
nicating with friends (Ellison, Steinfield, & 2011; Pempek et al., 2009; Ross et al., 2009;
Lampe, 2007) and maintaining social connec- Underwood, Kerlin, & Farrington-Flint, 2011).
tions (Subrahmanyam, Reich, Waechter, & This method for operationalizing Facebook use
Espinoza, 2008). Social behavior and interper- suffers from a lack of methodological rigor, as
sonal relationships are essential to the develop- these authors did not provide detailed documen-
ment of emerging adults, and social media ap- tation of how they progressed through formal
pears to play an increasing role in these areas procedures for scale development and valida-
(Reich, 2010). tion. Few authors developing their own scale
used exploratory factor analysis (EFA), and
Measuring Social Media Use none completed a confirmatory factor analysis
(CFA) or provided detailed psychometric statis-
Perhaps because the concept of social media tics such as testretest reliability coefficient es-
use is relatively new and the rise to prominence timates. Other authors provided only vaguely
of its most popular manifestations is even defined measures (e.g., Kujath, 2011) and did
40 JENKINS-GUARNIERI, WRIGHT, AND JOHNSON

Note. Jenkins-Guarnieri, Wright, & Johnson (2012) refers to the present studys data. Cronbachs alpha coefficient; SMUIS Social Media Use Integration Scale; SIEC
Social Integration and Emotional Connection subscale; ISR Integration into Social Routines subscale; OSF Online Sociability Functions subscale; SMUIS total score testretest

Two factual items and six items using the 5-point Likert scale. b Indicates subscale. c Items used variable response formats that included: yes/no, 5-point Likert scale, and
not offer detailed psychometrics (e.g., Kwon &

EFA and CFA


Methods
Wen, 2010; Peluchette & Karl, 2010), making
evaluations of their instruments difficult.
On the other hand, some researchers have

EFA

EFA
used measures of Facebook use based on pre-
viously published scales (see Table 1 for over-

Number of items

Attitudesb 7
view). One commonly used measure created by

SMUIS 10
SIECb 6

OSFb 5
Ellison et al. (2007) is the Facebook Use Inten-

ISRb 4
FIS 8
sity scale, which has been used by a number of
researchers (e.g., Orr et al., 2009; Valenzuela,
Park, & Kee, 2009). Although developed for the
purposes of their study, Ellison and colleagues

Attitudesb .85
Testretest .80
(2007) created their scale using college student

SMUIS .91
SIECb .89
Reliability

OSFb .74
ISRb .83
participants, and reported item means and inter-
nal consistency reliability estimates from these

.83
data. However, they did not report any addi-
tional analyses in developing this scale, and no
additional psychometric evidence has been pub-

a
lished on data collected using this scale. Addi-

5-point Likert scale

6-point Likert scale


Response format
tionally, its usage is far from uniform across the
literature; for example, Kalpidou, Costin, and
Morris (2011) treated the first two items assess-

Variablec
ing behavioral frequency of usage as two dis-
tinct variables and the mean of the six remain-

over a 3-week interval; EFA exploratory factor analysis; CFA confirmatory factor analysis.
ing items assessing emotional connection to
Facebook Intensity

active Facebook use as another variable. Thus,


subscale names

Questionnaire
Measure and

even the popular Facebook Use Intensity scale


Scale (FIS)

suffers from a lack of methodological rigor in


Attitudesb
Facebook

development, strong psychometric evidence,


SMUIS
SIECb

and administration standardization, suggesting OSFb


ISRb

multiple-choice formats that ranged from a 4-item to a 9-item multiple choice.


it may be a weak measure of Facebook use.
Another scale that has been used by others
Undergraduate college

Undergraduate college

Undergraduate college

(e.g., Ryan & Xenos, 2011) is the Facebook


Questionnaire developed by Ross et al. (2009)
Sample

during the course of their study on Facebook


students

students

students

use and personality factors (see Table 1). This


28-item survey included the six attitudinal items
adapted from Ellison et al. (2007) and an addi-
tional item created by the authors, along with a
number of items assessing behavioral usage of
Mean

20.1

18.4

21.7
age

Facebook and its various features (e.g., the


Tag function and Wall posts). Items were
Female 188

Female 389

rated by participants on a mixture of Likert-type


Female 82
Participants

Male 162
Male 98

Male 15

rating scales and dichotomous responses (e.g.,


n 286

n 552
Measures of Facebook Use

n 97

yes/no). Ross and colleagues (2009) performed


principal components factor analysis with a va-
rimax rotation to yield latent variables instead
Ellison et al., 2007

of using single indicators in their analyses. This


Authors and year

Jenkins-Guarnieri,

Ross et al., 2009


Johnson, 2012

produced two distinct factors that they called


Attitudes ( .85), which comprised the six
Wright, &

items adapted from Ellison et al. (2007), and


Table 1

Online Sociability Functions ( .74), which


comprised items assessing behavioral frequency
a
SOCIAL MEDIA USE INTEGRATION SCALE 41

of Facebook use. However, the varimax rotation (Subrahmanyam et al., 2008). As recent meth-
assumes an orthogonal/uncorrelated relation- ods of assessing Facebook use appear to suffer
ship between factors, and using this rotation from a lack of methodological rigor, the present
with correlated factors may yield overestimated study developed the Social Media Use Integra-
loadings as well as inappropriate retention and tion Scale (SMUIS), a measure of social media
rejection of items (Worthington & Whittaker, use integration with detailed psychometric evi-
2006), which risks a solution that is an artifact dence for an emerging adult population. In ad-
of the methods used. Thus, this measure did not dition, we presented preliminary evidence for
benefit from strong psychometric analyses, and the internal structure as well as convergent and
future research must determine whether similar discriminant validity of the SMUIS. This instru-
results are an artifact of the specific methods ment was designed to assess Facebook use in-
used. tegration but was intended to be flexible enough
In addition, the Ross et al. (2009) scale es- to be adapted to other SNS and online social
sentially used Ellison and colleagues (2007) media.
Facebook Use Intensity scale items (which cap-
tured integration of usage into social behavior Methods
and the emotional connection to this use), and
developed behavioral frequency items that rep- Participants
resented a different latent construct. Although
these results present a possible measurement All participants were undergraduate students
scale of behavioral frequency specific to Face- between the ages of 17 and 25 years recruited
book.com, they do not present strong psycho- from a medium-sized (N 12,000) Rocky
metric evidence for the Ellison et al. (2007) Mountain region university. Using a conve-
scales data collected from a sample of college nience sampling method, we invited all 3,022
students. In this way, the novel part of this scale first-year students to participate in this research
involved behavioral frequency items only spe- through their university e-mail accounts and
cific to one SNS (i.e., Facebook), and did not directed them through a hyperlink to the studys
present a measure of the degree to which social online survey. A total of 627 participants re-
media use is integrated into overall social be- sponded to the invitation e-mail and began the
havior. Thus, this measure also fails to present online survey (20.74% response rate), and the
strong psychometric evidence for its use with 616 (98%) who reported using a Facebook ac-
emerging adults. count were included in this study. The mean age
of participants was 18.42 (standard deviation
Study Rationale [SD] 0.996) years, and 71.1% reported being
female, with 72.6% identifying as Caucasian,
Given the call for further research on the 11% multiracial, 9.6% Hispanic/Latino, 4.5%
potential for Internet use to influence interper- African American, 1.3% Asian, 0.5% Native
sonal relationships and social skills (Thayer & American, and 0.3% Pacific Islander. From this
Ray, 2006), researchers have focused on the total sample (N 616), 482 participants chose
integration of social media into the daily lives to enter their e-mail address after completing
and social behavior of users (Steinfield et al., the survey and were e-mailed again 3 weeks
2008). As Internet use increases in emerging later with an invitation to retake the same 22-
adults (Zickurh, 2010), researchers have urged item scale for use with calculating testretest
further investigations into social media use reliability. Ninety-five of these participants re-
(Brown, 2006; Ross et al., 2009), as well as the sponded within 5 days to complete all the items
role of this use in social behavior (Raacke, a second time, and the demographics of this
2008), especially Facebook.com. For the cur- sample did not differ significantly from the
rent study, we chose to use a college sample larger sample.
because college students between the ages of 18
and 25 years are described as emerging adults Procedures
(Arnett, 2000), and emerging adults often use
Facebook to initiate and maintain relationships After receiving institutional review board ap-
(Ellison et al., 2007) and social connections proval, data were collected through a secure
42 JENKINS-GUARNIERI, WRIGHT, AND JOHNSON

Web-based survey site hosted by SurveyMon- yield a scale composed of 22 potential items.
key.com (SurveyMonkey, 2011); the first 25 Items 15, 22, and 9 were removed because the
participants who completed the survey earned a content of these statements was redundant with
small Amazon.com gift card ($2) as an incen- items 13, 10, and 4, respectively, and the items
tive. We randomly split the full sample in half did not add to the psychometric strength of the
to form a calibration sample (n 308) for an measure. The first three items focused on fre-
EFA and a hold-out sample (n 308) for a quency of social media use, such as average
CFA to test the proposed scales fit to the data. minutes spent using these services per day, and
The calibration sample had similar demo- the remaining items were created to capture a
graphic characteristics as the full sample users integration of the site into social behav-
(70.1% female, mean age of 18.4 [SD 1.05] iors and routines, as well as emotional invest-
years, 72.1% Caucasian, 11.7% multiple races, ment in the sites use. The instructions for the
8.4% Hispanic, 4.9% African American, 1.6% participants stated Please indicate how much
Asian, 0.6% Pacific Islander, and 0.6% Native you agree or disagree with the following state-
American). The hold-out CFA sample also had ments, and a Likert-type response scale was
similar demographics to the full sample (72.4% created for participants to indicate their level of
female, mean age of 18.37 [SD 0.938] years, agreement or disagreement with each items
73.4% Caucasian, 10.5% multiple races, 10.1% statement using anchors ranging from 1
Hispanic, 4.2% African American, 1% Asian, (Strongly disagree) to 6 (Strongly agree),
and 0.3% Native American). The testretest with numbers two through five listed sequen-
sample of 95 reported being 67.4% female, had tially and spaced evenly in between. EFAs re-
a mean age of 18.4 (SD 0.795) years, and sulted in a 13-item two-factor solution (see
identified as 69% Caucasian, 9.5% Hispanic, Table 2), and subsequent scale development
9.5% multiple heritages, and 1.1% Asian. analyses (i.e., CFAs) resulted in a final 10-item
two-factor scale (see Table 3 for scale items),
Measures with higher scores reflecting more engaged use
and integration of social media. For our data,
Social media use integration. A pool of the Cronbachs alpha coefficient for total scale
potential scale items for the SMUIS was created scores was .914, and for scores on the sub-
through collaboration with two psychologists scale 1 and 2 were, .893 and .828, respectively.
familiar with current research on social media Facebook use intensity. Ellison et al.
use and emerging adult populations. In addition, (2007) created an 8-item scale called Facebook
we adapted items from previously published Use Intensity. The first two items asked about
scales such as Ellison et al.s (2007) and Ross et the average number of minutes spent actively
al.s (2009) measures. Our final scale had one using Facebook per week and about the number
item (i.e., using Facebook is part of my every- of people connected to a users account (i.e.,
day routine) related to social routines that over- Facebook friends), on which participants could
lapped with two items on Ellison et al.s scale elect one of the nine custom answers (e.g., 0
that included: (1) Facebook has become part of 10 or less, 1 1150). The remaining six
my daily routine and (2) Facebook is part of items asked users about their emotional connec-
my everyday activity, and this last item was tion to the site and its role in their social rela-
also included on Ross et al.s scale. Finally, an tionships, and participants used a Likert-type
informal focus group that consisted of three scale ranging from 1 (Strongly Disagree) to 5
psychologists, two doctoral students, and three (Strongly Agree) to indicate their level of
undergraduate students was conducted to deter- agreement or disagreement. Ellison et al. (2007)
mine the appropriateness and usefulness of the reported a Cronbachs alpha estimate of .83 for
proposed items (DeVellis, 2003) by comparing data obtained from a sample of emerging adults
the new items with previously published items in college. This scale was administered in our
(e.g., Ellison et al., 2007; Ross et al., 2009); no study to provide convergent validity evidence
new original items were created from this for the SMUIS.
group. These procedures yielded an initial 34- Five-factor model personality traits. The
item pool that then underwent a process of Conscientiousness and Agreeableness subscales
revisions by the workgroup of collaborators to of the Big Five Inventory (BFI; John, Naumann,
SOCIAL MEDIA USE INTEGRATION SCALE 43

Table 2
Scale Items and Pattern Coefficients for 13-Item Two-Factor Model
Two-factor solution
Item Item text Factor 1 Factor 2
8 I get upset when I cant log on to Facebook 0.886 0.147
10 I prefer to communicate with others mainly through Facebook 0.873 0.172
13 Facebook plays an important role in my social relationships 0.697 0.092
6 I would like it if everyone used Facebook to communicate 0.685 0.137
5 I feel disconnected from friends when I have not logged into Facebook 0.683 0.084
7 I would be disappointed if I could not use Facebook at all 0.630 0.190
18 I share many of my day to day activities through Facebook 0.594 0.075
12 I check Facebook immediately when I am alerted of new activity on my account 0.501 0.153
2 On average, how many days per week do you use Facebook? 0.163 0.839
14 Using Facebook is part of my everyday routine 0.067 0.786
4 I enjoy checking my Facebook account 0.108 0.703
11 I dont like to use Facebook (r) 0.020 0.597
17 I respond to content that others share using Facebook 0.319 0.491
Note. n 279. Item 2 was standardized owing to a different scale being used for the item that consisted of the numbers
1 through 7. Factor 1 social integration and emotional connection; factor 2 integration into social routines. (r) item
11 reverse coded. Bold items were retained for respective factors.

& Soto, 2008) were developed to measure those lated to Facebook use, and were chosen for use
two constructs from the Five-Factor Model. with establishing discriminant validity evidence
These subscales each consist of nine items on for our new scale.
which participants indicated their disagreement
or agreement with item statements on a Likert- Data Analyses
type rating scale ranging from 1 (Disagree
strongly) to 5 (Agree strongly). John et al. Using the data collected from 308 partici-
(2008) previously found adequate reliability es- pants in the calibration sample, all 19 proposed
timates with U.S. adults, ranging from 0.75 to items were submitted to an EFA using maxi-
0.90, and also presented evidence for conver- mum likelihood extraction method and a pro-
gent, discriminant, and concurrent validity. Pre- max (oblique) rotation (k 4) with Kaiser
vious research has found that Agreeableness normalization using IBM SPSS Statistics (re-
(Wilson, Fornasier, & White, 2010) and Con- lease 19). Mundfrom, Shaw, and Ke (2005)
scientiousness (Ross et al., 2009) were unre- recommended a minimum sample size of 200

Table 3
The Final 10-Item Scale and Descriptive Statistics Using the Total Sample
Item Item text Mean SD ITC
a
5 I feel disconnected from friends when I have not logged into Facebook 3.24 2.56 0.691
6a I would like it if everyone used Facebook to communicate 3.30 1.52 0.758
7a I would be disappointed if I could not use Facebook at all 3.61 1.69 0.750
8a I get upset when I cant log on to Facebook 2.70 1.55 0.701
10a I prefer to communicate with others mainly through Facebook 2.58 1.39 0.642
13a Facebook plays an important role in my social relationships 2.96 1.42 0.709
4b I enjoy checking my Facebook account 4.70 1.24 0.692
11b I dont like to use Facebook (r) 4.99 1.22 0.533
14b Using Facebook is part of my everyday routine 4.07 1.50 0.683
17b I respond to content that others share using Facebook 3.84 1.42 0.670
Note. n 552. SD standard deviation; ITC corrected item-total correlation. Scale range for items: 1 strongly
disagree to 6 strongly agree. (r) item 11 reverse coded.
a
Social Integration and Emotional Connection subscale. b Integration into Social Routines subscale.
44 JENKINS-GUARNIERI, WRIGHT, AND JOHNSON

for EFA with a variables-to-factors ratio of six, Bonferroni adjustment: .05/2 .025). Skew
wide communality, and a coefficient of congru- and kurtosis statistics and plots of standardized
ence K .98 (excellent criteria); our sample residuals were examined to evaluate whether
size met these minimum requirements. Items 1 these data met the assumptions of homoscedas-
through 3 were standardized before being in- ticity, independence of observations, and mul-
cluded in analyses, given the nature of their tivariate normality underlying this analysis.
content and large ranges and variances. Multi-
ple criteria were used to determine retention of Results
factors, including eigenvalues 1, a visual
analysis of the produced Scree plot, and inter- There was no evidence of collinearity in the
pretability of the factor solutions. Factor analy- data, and all items displayed skew and kurtosis
sis was conducted over a number of iterations, values within acceptable ranges. Both Bartletts
and items were removed if they displayed a Test of Sphericity (p .001) and the Kaiser
pattern coefficient of .45 or if they loaded on MeyerOlkin measure (KMO .920) sup-
more than one factor with a pattern coefficient ported the factorability of these data (Tabach-
of .32 (Tabachnick & Fidell, 2001). nick & Fidell, 2001). Using the calibration
Based on these results, we followed a Model sample, the 19 items were subjected to an EFA
Generating (MG) approach (Jreskog, 1993) to with data from 268 participants (40 participants
Structural Equation Modeling (SEM) in con- were excluded using listwise deletion owing to
ducting a CFA using EQS software (release 6.1) missing data from omitted responses), and three
to evaluate the fit of the observed indicators factors were retained using the aforementioned
selected by the EFA (the a priori model) to the criteria after the first iteration. Items 20 and 21
data on the same scale items from the separate loaded on their own factor, and were removed
hold-out sample. Following this method, the because of conceptually different item content
hypothesized models fit to the data was first concerning social media uses interference in
tested, and then, the results from these analyses daily functioning, and because they grouped
were used to respecify the model to yield a into a factor of only two items (Worthington &
statistically plausible and practically meaning- Whittaker, 2006). In addition, items 1, 3, 16,
ful model (Byrne, 2001); this approach matches and 19 were removed because of low pattern
the purpose of the current study in developing a coefficient loadings (.45). With data from 279
scale from novel items that adequately measures participants available, the remaining items were
the construct of social media use integration. subjected to a second iteration of EFA, which
The hold-out sample meets the minimum size produced a clear and interpretable rotated factor
requirements of 200 for most applications of structure: a 13-item solution with two factors
SEM (Kline, 2011; Weston & Gore, 2006), al- that we named SIEC (factor 1) and ISR
though greater numbers are always recom- (factor 2). Factor 1 had an eigenvalue of 6.738
mended. Following Klines (2011) guidelines in and accounted for 51.833% of the variance,
conducting SEM analyses, hypothesized models whereas factor 2 had an eigenvalue of 1.281 and
were identified using a unit loading constraint accounted for 9.854% of the variance; the pat-
for the first indicator of each factor, data were tern coefficients for the included items are dis-
screened for any evidence of collinearity by played in Table 2. Cronbachs alpha reliability
examining the squared multiple correlations for estimates calculated from this scales data were
values .9 (Tabachnick & Fidell, 2001), and .901 for factor 1, .843 for factor 2, and .921 for
skew and kurtosis statistics with absolute values the total scale; the interfactor correlation was
2 and 7 (Hoyle, 1995), respectively, were r .713.
examined for indications of deviation from the There were 273 participants from the hold-
assumption of normality underlying SEM. out sample after using listwise deletion to ad-
Owing to the imbalance in the gender and dress omitted responses. The 13-item two-factor
ethnicity of study participants, SMUIS subscale model of the SMUIS suggested by earlier EFA
mean scores were examined simultaneously for analyses was titled model 1 and subjected to a
potential differences by gender and ethnicity CFA using the MG approach to SEM; skew
(Caucasian and non-Caucasian) using a multi- (ranging from 1.444 to 0.747) and kurtosis
variate analysis of variance (alpha level with (ranging from 1.188 to 1.845) statistics fell
SOCIAL MEDIA USE INTEGRATION SCALE 45

within acceptable ranges for all items, and the cific item content for removed items 2, 12, and
largest standardized residual was .495. Maxi- 18 are listed in the footnote of Table 5. Model
mum likelihood (Bentler & Chih-Ping, 1987) 3s final 10-item two-factor structure demon-
estimation methods were used with the robust strated good fit to the data, with approximate fit
SatorraBentler 2 statistic (Satorra & Bentler, statistics falling within acceptable ranges:
1988) to avoid potential bias in the 2 statistic SRMR 0.041, RMSEA 0.075 (90% confi-
with non-normally distributed data (Hutchinson dence interval of 0.056 0.095), CFI 0.96,
& Olmos, 1998). The 2 statistic for exact and NNFI 0.95. Although the 2SB was statis-
model fit can be overly sensitive and biased by tically significant, suggesting inexact fit, the
a number of model characteristics (Kline, 2SB/df ratio of 2.53 was 3. In addition, stan-
2011), and thus, the revised criterion of 2/df dardized and unstandardized parameter esti-
3 (Iacobucci, 2010) was used. Based on pub- mates were all statistically significant and prac-
lished model evaluation guidelines, the follow- tically meaningful (see Table 6), suggesting
ing approximate fit statistics and cutoff criteria strong validity evidence for the internal struc-
were used: .08 for the standardized root mean ture of the SMUIS. A CFA conducted on the
square residual (SRMR), close to .06 for the root alternate one-factor 13-item model produced in-
mean square error of approximation (RMSEA) adequate fit, even after multiple model modifi-
(Hu & Bentler (1999), and .95 for the non- cations, which retained only nine items (SRMR
normed fit index (NNFI) and comparative fit 0.051, RMSEA 0.102, CFI 0.94, and
index (CFI) (Beauducel & Wittmann, 2005). NNFI 0.92, 2 127.284, and 2SB
These statistics were selected owing to consid- 102.951 [p .001]).
erations of sample size, model size, and the Convergent validity evidence was established
potential for non-normality in the data between the SMUIS and the Facebook Use In-
(Hutchinson & Olmos, 1998). tensity Scale. Using the final 10-item two-
Based on the CFA analysis, fit statistics for subscale SMUIS and the full data set (N 552
model 1 are shown in Table 5, and suggested after listwise deletion), both subscales and total
that model 1 did not demonstrate good fit to the mean scores demonstrated significant (p
data. Following the MG approach to SEM, .001) moderately large and positive correlations
modifications were made to model 1 based on with the Facebook Use Intensity Scales total
low R2 values and large univariate Lagrange mean score (items standardized), and are dis-
multiplier test statistics that indicated high error played in Table 4 along with Cronbachs reli-
covariances and cross-loadings for some items, ability estimates calculated from the final
which suggested that items 2 and 12 be removed scales data, which fell within acceptable
from model 1, resulting in model 2 that con- ranges. In addition, the two original items as-
sisted of 11 items. Analyses of fit statistics for sessing behavioral frequency of social media
model 2 also did not demonstrate a good fit to use (one and three) were significantly (p
the data and item 18 was removed, based on the .001) and positively correlated with the SMUIS
aforementioned procedures for model 1. Spe- SIEC subscale (r .303 and r .198) and the

Table 4
Descriptive Statistics and Bivariate Correlations for the Final 10-Item SMUIS and Other Measures
Measure 1 2 3 4 5 6 Mean SD
1. SMUIS factor 1 .893 3.06 1.23
2. SMUIS factor 2 0.705 .828 4.39 1.10
3. SMUIS total 0.959 0.878 .914 3.59 1.09
4. Facebook use intensitya 0.697 0.750 0.772 .852 0.009 0.71
5. Conscientiousness 0.022 0.079 0.017 0.031 .737 3.70 0.55
6. Agreeableness 0.006 0.106 0.038 0.038 0.387 .759 4.01 0.54
Note. n 552 using listwise deletion. SMUIS Social Media Use Integration Scale, mean scores used to calculate
correlations; factor 1 social integration and emotional connection; factor 2 integration into social routines.
a
Items standardized before any calculations owing to differing item ranges.

p .01.
46 JENKINS-GUARNIERI, WRIGHT, AND JOHNSON

Table 5
Overall Model Fit Statistics for Two-Factor Baseline and Nested Models
Model 2 SB2 df 2 difference RMSEA SRMR CFISB NNFISB
1 240.110 213.290 64 0.101 0.055 0.91 0.90
2 141.301 124.858 43 98.432 0.084 0.047 0.94 0.93
3 97.892 86.335 34 43.409 0.075 0.041 0.96 0.95
Note. n 273. SB2 SatorraBentler scaled chi-square statistic; RMSEA root mean square error of approximation;
SRMR standardized root mean square residual; CFI comparative fit index; NNFI non-normed fit index. Model 1
included all 13 items; model 2 excluded items 2 (on average, how many days per week do you use Facebook) and 12 (I check
Facebook immediately when I am alerted of new activity on my account); model 3 excluded items 2, 12, and 18 (I share
many of my day to day activities through Facebook).

p .01.

SMUIS ISR subscale (r .368 and r .196). significance to help provide very strong ev-
From a conceptual point of view, it is unlikely idence against the null (Royall, 1986, p. 313)
that individuals levels of conscientiousness and to reduce type I errors with our larger
or agreeableness would determine their Face- sample size.
book use. Therefore, evidence of discriminant Using data from the 95 participants who re-
validity was demonstrated using the BFI Con- took the final 10-item SMUIS 3 weeks after the
scientiousness and Agreeableness subscale first administration, the testretest reliability es-
mean scores, given previous research suggest- timate was r .804 for subscale 1s mean score
ing that they were not significantly related to and r .676 for subscale 2s mean score; the
Facebook use (e.g., Ross et al., 2009). These testretest correlation for the 10-item total mean
Conscientiousness and Agreeableness sub- score was r .803 (p .001 for all correlation
scales displayed nonsignificant correlations statistics). Results from a multivariate analysis
with the SMUIS-SIEC subscale (p .611 and of variance using all 616 participants investigat-
p .883, respectively) and SMUIS-ISR sub- ing potential differences in the two subscale
scale (p .062 and p .013, respectively); mean scores for gender and ethnicity (aggre-
results are presented in Table 4. In consider- gated into white and nonwhite owing to the low
ing the frequently published critical p values number of people of color) indicated no main
of .05 and .01 (Thompson, 2000), we decided effects (based on Wilks criterion) due to gender
to use a critical value of .01 to determine ( .999, F(2, 544) 0.129, p .879, 2p

Table 6
Unstandardized and Standardized Coefficients for the Final 10-Item Model
Latent construct/ Standardized
Item subscale B SE R2 error variances
8 1 1 .785 .617 0.619
10 1 0.770 .702 0.067 .493 0.712
13 1 0.889 .756 0.067 .571 0.655
6 1 1.001 .811 0.065 .658 0.585
5 1 0.935 .743 0.062 .552 0.670
7 1 1.067 .784 0.066 .615 0.620
14 2 1 .789 .623 0.614
4 2 0.849 .787 0.060 .620 0.617
11 2 0.693 .695 0.065 .483 0.719
17 2 0.919 .741 0.064 .549 0.672
F1F2 1.147 .831 0.119
Note. n 273. All unstandardized parameter estimates and error variances statistically significant at p .05. Robust
statistics used with robust ML estimation method. Latent construct/subscale 1 Social Integration and Emotional
Connection; latent construct/subscale 2 Integration into Social Routines. Critical ratios (C.R.) can be calculated by
dividing unstandardized estimates by standard errors. C.R. 1.96 are significant at p .05.
SOCIAL MEDIA USE INTEGRATION SCALE 47

.001) or ethnicity ( .995, F(2, 544) 1.419, ability was established over a 3-week interval.
p .243, 2p .005); therefore, there were no Validity evidence was demonstrated in a num-
gender or ethnic differences in the SMUIS sub- ber of different ways. Significant positive cor-
scale mean scores. relations were found between the SMUIS and
Facebook Use Intensity Scale, and nonsignifi-
Discussion cant correlations between the SMUIS and two
scales from the BFI provided evidence for con-
Previous measurement instruments for social vergent and discriminant validity. In addition,
media use in emerging adults have suffered both subscales scores exhibited significant pos-
from a lack of detailed psychometric support itive correlations with the two factual items
and validity evidence. Authors developing measuring minutes spent using the site and
scales of social media use have focused mainly number of Facebook friends, although at lower
on behavioral frequency of use (e.g., Ross et al., levels than correlations with the Facebook Use
2009), used inadequate measures such as single Intensity scale scores.
items (e.g., Baker & Oswald, 2010), or failed to It is important to note that three items from
present adequate psychometric and validity ev- the original scale development item pool fo-
idence for data collected (e.g., Carpenter et al., cused on more factual information about Face-
2011; Ellison et al., 2007). Ellison and col- book use, such as number of average daily
leagues have also suggested that measures of minutes spent using the site and number of
social media use should capture the integration Facebook friends. EFA analyses suggested that
of its use into the daily lives and social behavior these items were not strong measures of Face-
of users, as well as the emotional connection a book use integration, and that they were weaker
user develops to the media rather than simply than other items assessing the integration of
frequency-of-use estimates. To address these Facebook use into ones daily routines or emo-
limitations, we developed a brief 10-item scale tional attachment to this use. These results sug-
of social media use integration called the gested either that social media integration is a
SMUIS. separate construct from behavioral frequency of
use, or that behavioral measures may be weaker
Reliability and Validity Evidence for the means of operationalizing social media use and
SMUIS its integration. It is likely that although the
quantity of social media use (frequency/
The SMUIS was designed to assess engaged intensity) will vary depending on various fac-
use of a variety of social media in emerging tors in ones life, perceptions of social medias
adult populations. In its development, we fo- importance in ones life and the emotional con-
cused on the social media site Facebook.com; nection one forms with it will be more stable
however, the SMUIS was designed to be over time. This may be an important area of
adapted to other forms of social media by re- study for future research. Finally, an analysis of
placing the word Facebook in the item state- potential differences between the mean scores
ments with the name of other social media ser- for both subscales for different genders and
vices. It was developed using rigorous scale ethnicities suggested that no significant group
development methods (DeVellis, 2003), and differences were evident in the data. Thus, the
benefits from detailed information about the present study developed a novel measurement
scales psychometric properties. We conducted scale for assessing social media use integration,
an EFA followed by a CFA with a different with detailed psychometric data from a college
sample to establish validity evidence for the student sample, and preliminary validity evi-
instruments internal structure, and produced a dence for its use with this population and the
two-factor model with subscale 1 called SIEC social media service Facebook.com.
and subscale 2 called ISR (see Tables 2 and 3).
Scores from the 10-item two-subscale SMUIS Practical Implications
demonstrated adequate internal consistency es-
timates in this sample (0.914 for total scale Similar to the original efforts of Ellison et al.
scores, 0.893 for subscale 1 scores, 0.828 for (2007), our results suggested that measuring
subscale 2 scores), and strong testretest reli- social media use should focus on the integration
48 JENKINS-GUARNIERI, WRIGHT, AND JOHNSON

of a site into ones social behavior and routines, properties we have reported and to identify the
as well as the emotional connection users expe- measures appropriateness with noncollege stu-
rience with their use. In the present study, items dent samples. Additional work is needed to
reflecting this type of content were retained in amass validity evidence, including support for
the final model, whereas items assessing the convergent, discriminant, and concurrent valid-
more quantitative aspects of social media usage ity. In addition, given the prevalence of Face-
were removed because they did not fit well in book and other social media in the United
the model. These results are also consistent with States, further research must seek to determine
research by Reich (2010) and Raacke (2008), this scales appropriateness for use with other
which suggested that emerging adults are in- populations of social media users. Although our
creasingly incorporating social media into their scale was developed focusing exclusively on
daily lives and social behavior. As this level of Facebook, it is intended to be flexible enough to
incorporation increases, it is not surprising that be adapted for use with other types of social
items assessing integration and emotional con- media. However, given our focus on Facebook
nection would be stronger indicators than be- in development, more research is needed to
havioral frequency of use. Although previous
establish validity evidence when adapting the
scales of Facebook use have been published
SMUIS to other types of social media, such as
(e.g., Ross et al., 2009), currently no scales have
YouTube, Twitter, Vimeo, Flickr, and so forth.
benefited from rigorous methods in their devel-
opment, detailed psychometric support, or Similarly, when using this new scale, future
strong validity evidence. Our results support the researchers may want to examine the various
conclusions of Ellison et al. (2007), suggesting media platforms that are frequently used to ac-
that assessment of social media use would ben- cess social media (e.g., desktops, laptops,
efit from a focus on the integration of media into phones, and tablets) to test for differences
ones life and the emotional connection to this among platforms.
integration, rather than behavioral frequency or
intensity of use. Thus, the SMUIS can be used Conclusions
in research that calls for measuring social media
use integration in emerging adult populations. Social media is becoming increasingly more
prominent in our lives. More and more individ-
Limitations and Directions for Future uals, groups, and organizations are joining and
Research using social media networks as a primary means
for communication. Although previous research
Given the underrepresentation of racially
has attempted to understand the impact of social
and ethnically diverse students in this study,
media and characteristics of its early users,
these results may have limited generalizabil-
measurement issues and poor scale develop-
ity. However, the data collected with the
SMUIS did not exhibit gender or ethnicity ment have hindered progress in this emerging
differences for the subscale or full-scale mean area of research. Social media use is often de-
scores. The sample used was nonrandom, and fined and measured by looking at quantity
composed of voluntary participants, which indicators focusing on behavioral frequency of
may have produced significant selection bi- use and amount of interaction with social me-
ases. It is also possible that traits associated dia. However, our results suggest that the quan-
with study participation may have influenced tity of social media use is not as useful as
the manner in which they responded to the measuring how one integrates this use into
scale items and thus biased the results. Fur- social routines as well as ones emotional con-
thermore, data were collected only from par- nection thereto. The SMUIS provides a brief
ticipants who had a Facebook account, and instrument with psychometric and validity evi-
research on the use of our SMUIS scale with dence that is adaptable to a variety of social
other social media services is needed. media. With the SMUIS, researchers and clini-
To establish evidence for the validity and cians can better measure and operationalize so-
utility of any new scale, further research is cial media use integration in research focusing
needed to confirm the strong psychometric on emerging adults.
SOCIAL MEDIA USE INTEGRATION SCALE 49

References Computer-Mediated Communication, 15, 158


188. doi:10.1111/j.1083-6101.2009.01498.x
Arnett, J. J. (2000). Emerging adulthood: A theory of Hoyle, R. H. (1995). Structural equation modeling:
development from the late teens through the twen- Concepts, issues, and applications. Thousand
ties. American Psychologist, 55, 469 480. doi: Oaks, CA: Sage Publications. (Retrieved from
10.1037/0003-066X.55.5.469 www.csa.com)
Baker, L. R., & Oswald, D. L. (2010). Shyness and Hu, L., & Bentler, P. M. (1999). Cutoff criteria for
online social networking services. Journal of So- fit indexes in covariance structure analysis: Con-
cial and Personal Relationships, 27, 873 889. doi: ventional versus new alternatives. Structural
10.1177/0265407510375261 Equation Modeling, 6, 155. doi:10.1080/
Beauducel, A., & Wittmann, W. W. (2005). Simula- 10705519909540118
tion study on fit indexes in CFA based on data with Hutchinson, S. R., & Olmos, A. (1998). Behavior of
slightly distorted simple structure. Structural descriptive fit indexes in confirmatory factor anal-
Equation Modeling, 12, 4175. doi:10.1207/ ysis using ordered categorical data. Structural
s15328007sem1201_3 Equation Modeling, 5, 344 364. doi:10.1080/
Bentler, P. M., & Chih-Ping, C. (1987). Practical 10705519809540111
issues in structural modeling. Sociological Meth- Iacobucci, D. (2010). Structural equations modeling:
ods & Research, 16, 78 117.(Retrieved from Fit indices, sample size, and advanced topics.
EBSCOhost). Journal of Consumer Psychology, 20, 90 98. doi:
Boyd, D. M., & Ellison, N. B. (2007). Social network 10.1016/j.jcps.2009.09.003
sites: Definition, history, and scholarship. Journal John, O. P., Naumann, L. P., & Soto, C. J. (2008).
of Computer-Mediated Communication, 13, 210 Paradigm shift to the integrative big five trait tax-
230. doi:10.1111/j.1083-6101.2007.00393.x onomy: History, measurement, and conceptual is-
Brown, J. (2006). Emerging adults in a media- sues. In O. P. John, R. W. Robins & L. A. Pervin
saturated world. In J. Arnett, & J. Tanner (Eds.), (Eds.), Handbook of personality psychology: The-
Emerging adults in America: Coming of age in the ory and research (3rd ed., pp. 114 158). New
21st century (pp. 279 299). New York: American York, NY: Guilford Press.
Psychological Association. doi:10.1037/11381- Jreskog, K. G. (1993). Testing structural equation
012 models. In K. A. Bollen & J. S. Long (Eds.),
Byrne, B. M. (2001). Structural equation modeling Testing structural equation models (pp. 294 316).
with AMOS: Basic concepts, applications, and Newbury Park, CA: Sage.
programming. Mahwah, NJ: Lawrence Erlbaum Kalpidou, M., Costin, D., & Morris, J. (2011). The
Associates Publishers. relationship between Facebook and the well- being
Carpenter, J. M., Green, M. C., & LaFlam, J. (2011). of undergraduate college students. Cyberpsychol-
People or profiles: Individual differences in online ogy, Behavior, and Social Networking, 14, 183
social networking use. Personality and Individual 189. doi:10.1089/cyber.2010.0061
Differences, 50, 538 541. Kline, R. B. (2011). Principles and practice of struc-
Cheung, C. M. K., Chiu, P., & Lee, M. K. O. (2011). tural equation modeling (3rd ed.). New York, NY:
Online social networks: Why do students use Fa- Guilford Press. (Retrieved from www.csa.com)
cebook? Computers in Human Behavior, 27, Kujath, C. L. (2011). Facebook and MySpace: Com-
13371343. plement or substitute for face-to-face interaction?
Correa, T., Hinsley, A. W., & de Ziga, H. G. Cyberpsychology, Behavior, and Social Network-
(2010). Who interacts on the web? The intersection ing, 14, 7578. doi:10.1089/cyber.2009.0311
of users personality and social media use. Com- Kwon, O., & Wen, Y. (2010). An empirical study of
puters in Human Behavior, 26, 247253. doi: the factors affecting social network service use.
10.1016/j.chb.2009.09.003 Computers in Human Behavior, 26, 254 263. doi:
DeVellis, R. F. (2003). Scale development: Theory 10.1016/j.chb.2009.04.011
and application (2nd ed.). Thousand Oaks, CA: Litt, D. M., & Stock, M. L. (2011). Adolescent alco-
Sage Publications. hol-related risk cognitions: The roles of social
Ellison, N. B., Steinfield, C., & Lampe, C. (2007). norms and social networking sites. Psychology of
The benefits of Facebook friends: social capital Addictive Behaviors, 25, 708 713. doi:10.1037/
and college students use of online social network a0024226
sites. Journal of Computer-Mediated Communica- Manago, A. M., Graham, M. B., Greenfield, P. M., &
tion, 12, 11431168. doi:10.1111/j.1083-6101 Salimkhan, G. (2008). Self-presentation and gen-
.2007.00367.x der on MySpace. Journal of Applied Developmen-
Grasmuck, S., Martin, J., & Zhao, S. (2009). Ethno- tal Psychology, 29, 446 458. doi:10.1016/j
racial identity displays on Facebook. Journal of .appdev.2008.07.001
50 JENKINS-GUARNIERI, WRIGHT, AND JOHNSON

Mundfrom, D. J., Shaw, D. G., & Ke, T. L. (2005). Applied Developmental Psychology, 29, 434 445.
Minimum sample size recommendations for con- doi:10.1016/j.appdev.2008.07.002
ducting factor analyses. International Journal of Subrahmanyam, K., Reich, S. M., Waechter, N., &
Testing, 5, 159 168. doi:10.1207/ Espinoza, G. (2008). Online and offline social net-
s15327574ijt0502_4 works: Use of social networking sites by emerging
Ong, E. Y. L., Ang, R. P., Ho, J. C. M., Lim, J. C. Y., adults. Journal of Applied Developmental Psychol-
Goh, D. H., Lee, C. S., & Chua, A. Y. K. (2011). ogy, 29, 420 433. doi:10.1016/j.appdev.2008.07
Narcissism, extraversion and adolescents self- .003
presentation on Facebook. Personality and Indi- SurveyMonkey. (2011). SurveyMonkeyAbout us.
vidual Differences, 50, 180 185. doi:10.1016/j Retrieved from http://www.surveymonkey.com/
.paid.2010.09.022 AboutUs.aspx
Orr, E. S., Sisic, M., Ross, C., Simmering, M. G., Tabachnick, B. G., & Fidell, L. S. (2001). Using
Arseneault, J. M., & Orr, R. R. (2009). The influ- multivariate statistics (4th ed.). Boston, MA: Al-
ence of shyness on the use of Facebook in an lyn & Bacon/Pearson Education. (Retrieved from
www.csa.com)
undergraduate sample. CyberPsychology & Be-
Thayer, S. E., & Ray, S. (2006). Online communica-
havior, 12, 337340. doi:10.1089/cpb.2008.0214
tion preferences across age, gender, and duration
Peluchette, J., & Karl, K. (2010). Examining stu-
of Internet use. CyberPsychology & Behavior, 9,
dents intended image on Facebook: What were 432 440. doi:10.1089/cpb.2006.9.432
they thinking?! Journal of Education for Busi- Thompson, B. (2000). AERA editorial policies re-
ness, 85, 30 37. doi:10.1080/08832320903217606 garding statistical significance testing: Three sug-
Pempek, T. A., Yermolayeva, Y. A., & Calvert, S. L. gested reforms. Educational Researcher, 25(2),
(2009). College students social networking expe- 26 30.
riences on Facebook. Journal of Applied Develop- Underwood, J. D. M., Kerlin, L., & Farrington-Flint,
mental Psychology, 30, 227238. doi:10.1016/j L. (2011). The lies we tell and what they say about
.appdev.2008.12.010 us: Using behavioural characteristics to explain
Raacke, J. (2008). MySpace and Facebook: Applying Facebook activity. Computers in Human Behavior,
the uses and gratifications theory to exploring 27, 16211626. doi:10.1016/j.chb.2011.01.012
friend-networking sites. CyberPsychology & Be- Valenzuela, S., Park, N., & Kee, K. F. (2009). Is
havior, 11, 169 174. doi:10.1089/cpb.2007.0056 there social capital in a social network site? Face-
Reich, S. M. (2010). Adolescents sense of commu- book use and college students life satisfaction,
nity on MySpace and Facebook: A mixed- meth- trust, and participation. Journal of Computer-
ods approach. Journal of Community Psychology, Mediated Communication, 14, 875901. doi:
38, 688 705. doi:10.1002/jcop.20389 10.1111/j.1083-6101.2009.01474.x
Ross, C., Orr, E. S., Sisic, M., Arseneault, J. M., Weston, R., & Gore, P. A., Jr. (2006). A brief guide
Simmering, M. G., & Orr, R. R. (2009). Personal- to structural equation modeling. The Counseling
ity and motivations associated with Facebook use. Psychologist, 34, 719 751. doi:10.1177/
Computers in Human Behavior, 25, 578 586. doi: 0011000006286345
10.1016/j.chb.2008.12.024 Wilson, K., Fornasier, S., & White, K. M. (2010).
Royall, R. M. (1986). The effect of sample size on Psychological predictors of young adults use of
the meaning of significance tests. The American social networking sites. Cyberpsychology, Behav-
Statistician, 40, 313315. ior, and Social Networking, 13, 173177. doi:
Ryan, T., & Xenos, S. (2011). Who uses Facebook? 10.1089/cyber.2009.0094
An investigation into the relationship between the Worthington, R. L., & Whittaker, T. A. (2006). Scale
big five, shyness, narcissism, loneliness, and Face- development research: A content analysis and rec-
book usage. Computers in Human Behavior, 27, ommendations for best practices. The Counseling
Psychologist, 34, 806 838. doi:10.1177/
1658 1664. doi:10.1016/j.chb.2011.02.004
0011000006288127
Satorra, A., & Bentler, P. M. (1988). Scaling correc-
Zickurh, K. (2010). Generations online in 2010. Re-
tions for chi-square statistics in covariance struc-
trieved from Pew Internet & American Life Project
ture analysis. Proceedings of the Economics and website http://pewresearch.org/pubs/1831/
Statistics Section, 308 313. Alexandria, VA: generations-online-2010.
American Statistical Association
Steinfield, C., Ellison, N. B., & Lampe, C. (2008). Received December 14, 2011
Social capital, self-esteem, and use of online social Revision received August 6, 2012
network sites: A longitudinal analysis. Journal of Accepted August 29, 2012

Potrebbero piacerti anche