The role of emotion in the learning process:Comparisons between online and face-to-face learning settings
Gwen C.Marchand ⁎,Antonio P.Gutierrez
University of Nevada,Las Vegas,Department of Educational Research,Cognition,and Development,4505S.Maryland Parkway,Box 453003,Las Vegas,NV 89154-1658,United States
a b s t r a c t
a r t i c l e i n f o Available online 17October 2011Keywords:
Achievement emotions Motivation
Online and face-to-face learning Learning strategies
As the presence of online and hybrid coursework at institutions of higher education has increased,so too has interest among educators and scholars in understanding personal and contextual factors that predict success in different types of learning environments.The purpose of the present study was to examine the relations among temporally-ordered variables,including beginning-semester self-ef ficacy,
utility value,and relevance of instruction,mid-semester emotions (hope,frustration,and anxiety),and end-of-semester learning strate-gies in a sample of 291graduate students (N =219for the traditional education group and N =72for the dis-tance education group)enrolled in an introductory research methods course.Multigroup path analyses were performed to test the equality of path coef ficients among the two groups.Results demonstrate that the groups differed with respect to several paths,including the paths from:extrinsic utility value to anxiety and to hope;relevance to hope;and frustration and anxiety to learning strategies.Implications for research,theory,and practice are discussed.
©2011Elsevier Inc.All rights reserved.
1.Introduction
As the presence of online and hybrid coursework at institutions of higher education has increased,so too has interest among educators and scholars in understanding personal and contextual factors that in fluence student choice in learning medium and in turn,which fac-tors predict success in different types of learning environments.Researchers have discovered that student preference for online learn-ing environments and engagement with online material is related to student self-ef ficacy for on
line learning (Artino,2010;Clayton,Blum-berg,&Auld,2010),self-ef ficacy for computer use and for self-regulation (Spence &Usher,2007),and learning orientations that in-volve independence and organization (Hoskins &van Hooff,2005).Students may view online environments as offering bene fits with re-spect to collaboration,self-regulated learning,and information seek-ing over those offered in more traditional classrooms (Lee &Tsai,2011)and some research suggests that students participating in on-line learning environments evidence greater achievement than their peers in face-to-face classes (e.g.,Lim,Kim,Chen,&Ryder,2008).
Empirical research is only beginning to surface about the motiva-tional and emotional processes which unfold between enrollment choice and eventual course outcomes that might be similar or differ-ent for students in online compared to traditional face-to-face courses.Much research on online environments has focused on
course design and instructional strategies that in fluence student par-ticipation and performance (see Osborne,Kriese,Tobey,&Johnson,2009for discussion).Researchers concerned with differences be-tween online and traditional classrooms and how students experience blends of the two also highlight the importance of student perceptions of the learning environment and student motivation during the learn-ing process in predicting positive learning strategies and outcomes (e.g.,Ginns &Elli
s,2007;Lee &Tsai,2011).One other aspect of student experience that is just beginning to enter the research dialog about online learning is student emotion.
During the past decade,emotion has emerged as a vital element of the learning process but many questions remain about emotion in edu-cation (Pekrun,2005).Research has identi fied both classes of emotion and speci fic discrete emotions as predictive of student academic out-comes with a range of student populations (e.g.,Ainley,2006;Goetz et al.,2012;Linnenbrink-Garcia,Rogat,&Koskey,2011).Further,empir-ical evidence supports the theoretical notion that one way student emotional experience in fluences academic outcomes is as a conduit for a range of personal and contextual variables (Artino,La Rochelle,&Durning,2010;Daniels et al.,2009).The majority of research on the role of emotions in academic learning has centered on traditional,brick-and-mortar classroom learning situations.What is less known is how emotions function in online learning environments and whether the predictors and outcomes associated with academic emotions are similar or different from traditional classroom environments.Thus,the purpose of this study is to explore the role of emotion in the learning process by investigating whether relations among motivational factors,emotions,and academic learning strategies are consistent or differ for
Internet and Higher Education 15(2012)150–160
⁎Corresponding author.Tel.:+17028954303;fax:+17028951658.E-mail address:Gwen.Marchand@unlv.edu (G.C.
Marchand).1096-7516/$–see front matter ©2011Elsevier Inc.All rights reserved.doi:
10.1016/j.iheduc.2011.10.001
Contents lists available at SciVerse ScienceDirect
Internet and Higher Education
students learning in different modalities in a graduate-level research methods course.
2.Relevant literature and theoretical framework
Wosnitza and Volet(2005)called for more research on the role of emotions on the learning process in online environments,citing theo-retical and empirical work in general education as evidence of the im-portance of understanding the origin,direction,and impact of emotions in learning.Responding to that call,the literature on emotions in online learning environments has grown in recent years.Research has focused on in-depth description of both the arousal and expression of emotion during online learning tasks,often related to collaboration, taking into account the role of social partners and the features of the task in contributing to emotional ,Jarvenoja&Jarvela, 2005;Smith,2008;Wosnitza&Volet,2005;Zembylas,2008).Another emerging topic reports instructional
techniques and course design that can be used to enhance student emotional experiences and engage-ment with online ,Michinov&Michinov,2008;Shank, 2009).A third literature strand has quantitatively considered emotion as a personal factor that influences student learning and performance (e.g.,Artino,2010;Artino&Stephens,2009).This line of research is sit-uated within social-cognitive views of self-regulated learning and con-siders achievement emotions as a crucial aspect of the learning process, one which involves a complex interplay between personal and contex-tual factors(Pekrun,2006).The present study seeks to complement the last area of research.
2.1.Emotion in the learning process
The control-value theory of achievement emotions posits that stu-dents'motivational beliefs,perceptions of their learning environ-ment,cognitive quality,and other environmental factors influence students'control and value appraisals of academic situations,which in turn predict student emotions and eventual learning and achieve-ment outcomes.The relationships among the elements of this model are thought to be dynamic and reciprocal(Pekrun,2006).Adap-tations of this model,in combination with theory from self-regulated learning,have been used in research on online learning environments. The model most commonly applied in online research is a social-cognitive model of self-regulated learning that includes personal fac-tors,consisting of motivational beliefs and achievem
ent emotions,pre-dicting personal behaviors related to cognitive strategy use,and academic outcomes(Artino,2009a;Artino,2009b).
Much of what is known about achievement emotions stems from research with students participating in traditional educational set-tings.Pekrun and his colleagues have identified enjoyment,hope, pride,relief,anxiety,shame,hopelessness,anger/frustration,and boredom as commonly occurring academic emotions in undergraduate student populations across class,learning,and test-related situations (Pekrun,Elliot,&Maier,2009;Pekrun,Goetz,Frenzel,Barchfeld,& Perry,2011;Pekrun,Goetz,Titz,&Perry,2002).Positive emotions, such as enjoyment,hope,and pride,have been positively associated with intrinsic motivation,effort,self-regulation,and more sophisticated learning strategies(Pekrun et al.,2011),whereas negative emotions such as anger/frustration,shame,anxiety,and boredom have been asso-ciated with reduced effort,lower performance,increased external regu-lation,and decreased self-regulated learning strategies(Artino,2009b; Daniels et al.,2009;Pekrun et al.,2009).
There is limited research on discrete emotions in online learning contexts.Discrete emotions that have been investigated with respect to online learning in higher education include anger/frustration, boredom,and enjoyment.Artino and Stephens(2009)found that when frustration and boredom were p
aired with self-efficacy and task value to create adaptive and maladaptive motivation-emotion profiles,students with more adaptive profiles reported higher levels of self-regulated learning strategies,greater course satisfaction and performance,and motivation to enroll in future online courses. Other research with online students indicated that boredom and frus-tration were negatively related to course satisfaction and continuing motivation;however,they were unrelated to the use of elaboration as a learning strategy.Additionally,boredom negatively predicted metacognition whereas the relationship was positive for frustration (Artino,2009b).In other research,the three discrete emotions did not distinguish between students who reported preferring online or face-to-face courses(Artino,2010),nor were students taking online courses in their area of professional core more or less likely to expe-rience negative emotions than non-core students(Artino,2009a). This emerging literature has not yet tested temporally-ordered pro-cess models of self-regulated learning,such as those outlined in the literature on traditional classroom ,Artino et al.,2010; Daniels et al.,2009),which include emotion as an outcome of motiva-tional perceptions and an antecedent of learning strategies.The litera-ture on the role of emotions in online higher education environments lacks depth in two areas addressed by the present study(1)under-standing how emotionfits into the process of learning,and(2)investi-gation of other discrete emotions experienced during online courses, such as hope and anxiety.
2.2.Antecedents and consequences of emotion in online learning
Emotions serve a variety of functions in the academic environ-ment,including promoting or undermining behavioral and cognitive engagement,self-regulation of learning activities,and achievement (Linnenbrink-Garcia&Pekrun,2011).Research searching for the source of student achievement emotions has identified a complex and varied pool of proximal and distal antecedents of student emotion, (Ainley,Corrigan,&Richardson,2005;Assor,Kaplan,Kanat-Maymon, &Roth,2005;Jarvenoja&Jarvela,2005;Op't Eynde&Turner,2006; Pekrun et al.,2002;Ruthing et al.,2008).Yet,perhaps the most impor-tant implication of achievement emotions to emerge from the litera-ture is that achievement emotions are malleable,emerging from person–environment transactions,and may reflect academic adjust-ment(Pekrun et al.,2011;Schutz,Hong,Cross,&Osbon,2006). Thus,the study of the antecedents and consequences of achievement emotions in a variety of situations(such as testing or during course-work)and ,traditional or online courses)is crucial for un-derstanding how to create learning environments that can promote positive emotional experiences,which in turn enhance student learn-ing and performance.
In the present study,three predictors of course-related emotions were investigated:self-efficacy for le
arning research methods,per-ceived task value of research methods,and perceived relevance of in-struction.Self-efficacy,or student beliefs and expectations about their capabilities(Bandura,1977;Usher&Pajares,2008),has been considered one aspect of student control beliefs in the classroom(Pekrun et al., 2011).Higher self-efficacy has been consistently related to higher levels of positive emotions,such as hope,pride,and enjoyment and lower levels of negative emotions such as anger/frustration,shame,boredom, and Goetz,Cronjaeger,Frenzel,Ludtke,&Hall,2010; Pekrun et al.,2011).Research on self-efficacy in online learning has pre-dominantly investigated self-efficacy for computer use or online learn-ing(see Moos&Azevedo,2009for review).Self-efficacy for subject learning in relation to student emotion in online settings has been in-vestigated on a limited scale in mathematics(Spence&Usher,2007) and never in relation to learning research methods material at the grad-uate level.Thus,the present study included student perceptions of self-efficacy for learning research methods material as an antecedent to stu-dent course-related emotion.
The second antecedent of student emotion included in the present study was student task value.Prior research in online environments has demonstrated a positive association between task value and
151
G.C.Marchand,A.P.Gutierrez/Internet and Higher Education15(2012)150–160
learning strategy use(Artino,2009a;Artino,2009b;Artino&Stephens, 2006).Research has also demonstrated that task value shares an impor-tant association with academic emotions(Artino,2009b;Pekrun et al., 2002;Pekrun et al.,2011).Task value has been conceptualized in many different ways,but one major component of task value is utility value(Eccles&Wigfield,1995).Utility value is defined as“the value a task acquires because it is instrumental in reaching a variety of long-and short-range goals”(Eccles&Wigfield,1995,p.216).If students de-velop beliefs about the utility value of understanding research to im-prove their classroom practice,they should be more likely to have more positive emotional experiences in learning situations.
The theoretical frame for this study suggests that students'subjec-tive experiences of their learning environment,which may include features such as instructional support or task characteristics,should shape student academic emotions as well as their use of learning strategies(Artino,2009a;Artino,2009b;Pekrun,Frenzel,Goetz,& Perry,2007).Further,research operatin
g within a self-determination paradigm posits that not all academic activities are intrinsically moti-vating,but through providing external supports,such as a rationale as to why an activity is useful or relevant to students'lives,students may internalize the value of engaging in activities that may not be particular-ly compelling or intrinsically interesting(Deci&Ryan,1985;Reeve, Jang,Hardre,&Omura,2002;Ryan&Deci,2000).Fostering relevance is one such external support,which refers to actions that help students understand the connection of the academic material to their own inter-ests and goals(Assor,Kaplan,&Roth,2002).When students perceive instruction to be successful at fostering relevance,they are more likely to experience positive affect in learning situations and greater engage-ment(Assor et al.,2002;Deci,Eghrari,Patrick,&Leone,1994).In graduate-level education courses,instructional strategies to foster rele-vance may include connecting course material to real-world or practice applications.Perceived relevance of instruction has not yet been subject to empirical investigation in relation to emotion in online settings with students of higher education.
Student learning strategy use serves as the outcome of student emotional experience in the current investigation.Meaningful learn-ing strategy use is one element of self-regulatory behavior and is typ-ically conceptualized as a mediator between personal and contextual characteristics of students of hi
gher education and actual student per-formance(Pintrich,2004).Research on student emotions with re-spect to college-student learning has often focused on predicting academic achievement and performance(Daniels et al.,2009;Pekrun et al.,2009),at times bypassing student engagement,strategies,or at-titudes that should lead to better performance.Theoretically,positive emotions should lead to moreflexible strategy use(Isen,2008)and conversely negative emotions should lead to less meaningful strategy use,but that premise has been only minimally investigated in online environments(Artino,2009b;Artino&Stephens,2009).
Researchers have suggested that learning online“requires consider-able autonomy and self-direction”(Artino&Stephens,2009,p.572), perhaps even more so than in traditional learning environments as con-trol for learning is shifted from the teacher to the student(Hartley& Bendixen,2001).If positive student emotion contributes to more adap-tive patterns of self-regulated learning,as indicated by various Artino&Stephens,2009),then understanding the processes associated with student emotional experiences in online en-vironments becomes a critical task.As suggested by Schutz et al. (2006),emotional experiences involve person-environmental transac-tions that exist within particular activity settings.In education,those ac-tivity settings are the classroom,and Schutz and his colleagues recommend focusing inquiry on emotion in the activity setti
ng where the transactions occur.Therefore,if emotions exist as a result of per-son–environment transaction within certain activity settings,educators and scholars may pose questions as to whether there may be something different about the nature of the person-environmental transaction in online settings as compared to traditional settings.And if the nature of that transaction does differ,then it is possible that the relative influ-ence of antecedents of emotion or the strength of emotion in predicting academic outcomes may vary across settings.It is beyond the scope of the present study to investigate the nature of the transactions in the dif-ferent settings,so instead,the present study focuses on the relations among the antecedents and consequences of emotions across activity settings.
3.The present study
Graduate-level research methods coursework was selected as the context for the present study for two reasons.First,the research methods course is a required course for all students enrolled in a mas-ter's program in the college of education.Little research has been con-ducted with education majors about their experiences with research, but research with psychology students has noted that few students express interest,enthusiasm,or positive attitudes for taking research and statistics courses(Sizemore&Lewandowski,2009).Anecdotal in-formation from graduate students in research
methods courses would suggest that this population may also lack enthusiasm for this type of course,leading the researchers to believe that this course might be a rich venue for identifying a range of positive and negative emotions related to the course itself.Second,the research methods course is of-fered as a distance and a traditional face-to-face course,providing a venue for comparison between the two modalities.
There is no overwhelming theoretical or empirical justification to suggest that predictors and outcomes of academic emotions should differ for students in traditional or online formats.The lack of empirical evidence as to the learning process for distance education students led to the framing of this study as exploratory,posing research objectives rather than hypotheses.Thefirst objective of this study was to test a model based on the work of Pekrun(2006)and Artino(2009a, 2009b),investigating the role of emotion in the learning process for both traditional and online students in graduate-level research methods courses offered in a college of education.The conceptual model guiding the study design and analyses is shown in Fig.1.Accord-ing to Fig.1,extrinsic utility value,perceived relevance of instruction, and academic self-efficacy at the beginning of the semester are ante-cedents of mid-semester academic emotions of hope,frustration, and anxiety,respectively.The academic emotions are predictors of semester-end learning strategy use.The second objective of the study was to investi
gate whether there were differences in the model rela-tionships across students in online and face-to-face settings.
4.Method
4.1.Participants and design
Participants were291graduate students enrolled in both face-to-face and online sections of an introductory research methods course offered at a southwestern university located in a large urban area. The traditional face-to-face and online sections of the course were similar in many ways.An analysis of syllabi of courses offered during the study period indicated that both traditional and distance courses were taught by faculty members,adjunct professors with doctoral de-grees,and graduate assistants pursuing doctoral degrees.At least one faculty member taught both traditional and distance education sec-tions.Both formats had a cap of30students per section and were designed to be completed within a single semester.The distance courses were asynchronous in nature.The traditional course instruc-tional format consisted of combinations of lecture and group work, with the emphasis typically on lecture.The distance courses all in-cluded video or audio lectures and discussion activities,with varia-tions of other types of assignments.Participation and disc
ussion expectations were more formalized in distance education courses and more heavily weighted than in traditional formats.In both
152G.C.Marchand,A.P.Gutierrez/Internet and Higher Education15(2012)150–160
formats,one of four introductory textbooks was used and both types of classes evaluated students by a combination of tests and quizzes,a final project usually consisting of some kind of research article cri-tique,and participation or homework.
Students reported on a range of demographic and course history items,although not all students chose to respond to these questions.Fifty-three (18%)of participants were male and 132(45%)were fe-male (37%of participants did not respond).Participants'age ranged from 21to 68(M =33.5,SD =9.97),with over 75%seeking master's degrees in education with a variety of emphases (e.g.,educational leadership,sports education leadership,and curriculum and instruc-tion),and 113(39%)reporting that they currently work as teachers.The ethnic breakdown of the participants who reported this informa-tion was as follows:21(7%)Hispanic;120(41%)Caucasian;15(5%)African-American;21(7%)Asian-American/Paci fic Islander;and 8(3%)Other/Mixed (37%unreported).In terms of previous exposure to research methods or statistics,74(25%)participants reported hav-ing previou
s experience with research methods courses and 92(32%)had received previous training in statistics.Finally,166(57%)stu-dents reported having previously conducted research in some way.
The study design was correlational in nature.Students responded to questionnaires at three points during a single semester.All exoge-nous variables (i.e.,utility value,relevance,and self-ef ficacy)were taken from time 1,all emotion variables (i.e.,hope,frustration,and anxiety)were taken from time 2,and learning strategies (the out-come)were taken from time 3.Hence,although different time points were used,the study is not longitudinal because the focus was on dif-ferent,albeit temporally-ordered,variables across the semester.No attempt was made to draw causal inferences from these data,as the intent was to describe students'experiences in distance education and traditional/face-to-face graduate-level introductory research methods courses through a variety of measures.
4.2.Materials and instruments
4.2.1.Demographics
A 10-item demographic questionnaire was developed by the re-searchers to solicit information such as age,gender,degree and em-phasis,ethnicity,and prior exposure to research methods and statistics courses.In addition,questions were included that asked stu-dents to report previous experience cond
ucting research (e.g.,action
research,using the research literature to understand a problem in the classroom,and using research to enhance professional practice).4.2.2.Academic emotions
Pekrun,Goetz,and Perry's (2005)Achievement Emotions Ques-tionnaire (AEQ)was used to gauge students'achievement emotions in relation to the course.Only items related to emotions before and during class were included for the present study.Although the AEQ measures a range of positive and negative activating and deactivating emotions in line with the control value theory of achievement emo-tions (Pekrun et al.,2002),only the positive activating emotion of hope (8-item scale)and the negative emotions of anxiety (10-item scale)and anger/frustration (4-item scale)were used in this study.Sample items include,“I am full of hope (hope,before class),”“I feel anger welling up in me (frustration,during class),”and “I feel scared (anxiety,before class).”Students responded on a five point Likert scale from “strongly disagree ”(1)to “strongly agree ”(5).Instructions were provided to fit either setting.When responding to items related to before class,online students were instructed to “please indicate how you feel before signing in to access the material ”;when respond-ing to items related to during class,online students were instructed to “please indicate how you feel when you're working on the material ”.
The course-related AEQ has demonstrated acceptable reliability in previous studies,with alpha levels ranging from .84to .95for hope,.85to.91for frustration,and.89to.91for anxiety (Ouano,2011;Pekrun et al.,2005).Cronbach's alphas for the scales used in the present study,by setting,are located in Table 1.
4.2.3.Motivation
The motivational factors of self-ef ficacy and extrinsic utility value were also measured.Self-ef ficacy was measured using a 7-item scale adapted from the Motivated Strategies for Learning Questionnaire (MSLQ)developed by Pintrich,Smith,Garcia,and McKeachie (1991),which included items such as “I'm certain that I can master the skills taught in research methods this year,”and “I can do almost all the research methods coursework if I don't give up.”Self-ef ficacy items were answered on a 7-point Likert scale ranging from “Not at all true of me (1)”to “Very true of me (7)”.Internal consistency reli-ability coef ficients for the self-ef ficacy scale range from .89to .93in previous research (e.g.,Hadwin,Winne,Stockley,Nesbit,&Woszczyna,2001;Pintrich et al.,1991
).
Fig.1.The hypothesized path model for distance education and traditional groups.Key:a time 1variables (beginning of semester),b time 2variables (mid-semester),and c time 3variables (end-of-semester).
153
G.C.Marchand,A.P.Gutierrez /Internet and Higher Education 15(2012)150–160
The extrinsic utility value scale was adapted from the work of Eccles and Wigfield(1995)and measured students'value of the tasks.The2-items for this measure were“How useful is learning re-search methods for what you want to do after you graduate and go to work?”and“How useful is what you learn in research methods for your daily life outside school.”Possible responses ranged from “Useless(1)”to“Moderately useful(4)”to“Very useful(7)”.The inter-item correlation between these two items in the present study was moderate,r=.67.
4.2.4.Perceived relevance of instruction
An instrument designed for this study was developed to assess per-ceived relevance of instruction using items loosely adapted with the au-thors'permission from the work of Assor et al.(2002);Nix,Fraser,and Ledbetter(2005);and the Teacher as Social Context instrument(TASC; Belmont,Skinner,Wellborn,&Connell,1992).Sample items were,“The instructor provides examples of how research methods connects to real life,”and“The instructor discusses how I can use the information I'm learning in this course.”Students rated their agreement to these 8items on a7-point Likert scale ranging from“Not at all true(1)”to “Very true(7)”.
4.2.
5.Learning strategies
Greene,Miller,Crowson,Duke,and Akey's(2004)measure of meaningful strategy use was used to assess students'learning strate-gy use with the course material.This12-item instrument includes items such as,“Before a quiz or exam,I plan out how I will study,”“If I have trouble understanding something I go over it again until I understand it,”and“When I study I am aware of the ideas I have or have not understood.”Students responded to these items on afive point Likert scale from“strongly disagree”(1)to“strongly agree”
(5).Greene ported the Cronbach's alpha value for the measure to be.88.
4.3.Procedures
The research was approved by the university's institutional review board prior to the commencement of any data collection activities.A convenience sample of students was recruited from traditional and distance education settings of the graduate level introductory re-search methods course,and those who participated received course credit.All survey items were completed online via the Educational Psychology Department's Experiment Management System.Upon clicking on the research study link,students were taken to the elec-tronic informed-consent form where they read brief information a
bout the study.Subsequently,students were taken to the question-naire if they voluntarily agreed to participate in the study.Data were collected at three points during the semester:following two weeks of instruction,mid-way through the semester at week7,and two weeks prior to the end of the semester.Students had seven days to complete the study,albeit once they began the survey,stu-dents were expected to complete it in its entirety in one session
(i.e.,with no breaks).
4.4.Data analysis
Descriptive statistics,including internal consistency reliability coef-ficients for each measure,(see Table1)and correlations(see Table2) were conducted for all variables across pertinent time points.All data were screened for univariate and multivariate outliers according to the procedures outlined by Tabachnick and Fidell(2007)using the In-ternational Business Machine(IBM)Statistical Package for the Social Sciences(SPSS)Statistics19.No extreme outliers that would otherwise undermine the trustworthiness of the data were detected.Prior to data analysis,additional testing procedures detected several cases with missing data for both groups in the sample using EQS6.1(Bentler, 2005).The missing value analysis demonstrated that83cases(37.9%) in the traditional education grou
p and12cases(16.7%)in the distance education group had missing data.In order to verify that the missing data pattern was missing completely at random(MCAR),Little's MCARχ2statistics(Little&Rubin,1989;Schaeffer&Graham,2002) were requested from the missing values analysis.A significantχ, p b.05)would suggest that the pattern of missing data is not MCAR (i.e.,missing not at random[MNAR]),which poses a problem for inter-pretation of results because they may be biased due to systematic differ-ences in non-responses.However,the result of this test for the present data was non-significant for both groups,Little's MCARχ2(34)=32.55, p=.53(distance education group)and Little's MCARχ2(46)=60.05, p=.08(for the traditional education group),suggesting that the miss-ingness pattern in the data was MCAR.
In order to include all possible available data,maximum likelihood (ML)estimation(expectation maximization)was utilized to impute the missing data via EQS6.1,thereby yielding291available cases for analysis,72for the distance education group and219for the traditional education group.This ratio(approximately5:1in favor of traditional education courses)reflects the typical enrollment among face-to-face/ traditional and distance education courses at this university's College of Education.Furthermore,data were tested for univariate and multivariate assumptions,including multivariate normality(skewness and kurtosis),multicollinearity,and reproducibility of the correlation m
atrix via residual analysis using EQS6.1,in order to proceed with the path analysis with observed variables.Regarding multivariate nor-mality,Bentler(2005)stated that any data with Mardia's Normalized Estimate(MNE)>6.0is considered to be multivariate non-normal; the more the value of MNE differs from6,the greater the violation of multivariate normality.The data demonstrated moderate kurtosis for the traditional group(MNE for Multivariate Kurtosis=11.27);hence, the ML robust(MLR)statistics were requested and interpreted in lieu
Table1
Descriptive statistics and reliability coefficients for relevance,motivation,emotions,
and learning strategies by group.
Variables Distance a Traditional b
M SDαM SDα
Motivation
Self-efficacy 5.420.920.88 5.760.810.89
Utility value 4.88 1.090.70 4.98 1.260.82
Relevance 4.73 1.040.90 5.560.910.90
Emotions
Hope 3.490.500.80 3.690.610.87
Frustration 2.100.740.75 1.800.700.80
Anxiety 2.510.830.93 2.180.710.90
Learningcomparisons
Strategies 3.770.600.91 4.000.540.90
a N=72.
b N=219.
Table2
Zero-order correlations between motivational factors,relevance,emotions,and learn-
ing strategies by group.
Variable1234567
1.Utility value–.53⁎⁎.34⁎⁎.26⁎⁎−.33⁎⁎−.01.33⁎⁎
2.Relevance.63⁎⁎–.38⁎⁎.39⁎⁎−.36⁎⁎−.18⁎⁎.32⁎⁎
3.Self-efficacy.30⁎.32⁎⁎–.53⁎⁎−.39⁎⁎−.49⁎⁎.29⁎⁎
4.Hope.47⁎⁎.31⁎⁎.30⁎–−.56⁎⁎−.57⁎⁎.59⁎⁎
5.Frustration−.38⁎⁎−.51⁎⁎−.21−.57⁎⁎–.54⁎⁎−.42⁎⁎
6.Anxiety−.35⁎⁎−.36⁎⁎−.52⁎⁎−.61⁎⁎.59⁎⁎–−.09
7.Strategies.38⁎⁎.03.13.49⁎⁎−.09−.09–
Note.The traditional education group(N=219)correlation matrix is along the upper
diagonal while the matrix for the distance education group(N=72)is along the
lower diagonal.
⁎⁎p b.01(two-tailed).
⁎p b.05(two-tailed).
154G.C.Marchand,A.P.Gutierrez/Internet and Higher Education15(2012)150–160
of the ML normal distribution statistics.MLR procedures provide adjust-edfit ,S-Bχ2,*CFI,*NNFI,*IFI,and*RMSEA and its*CI90%) that correct for moderate-to-severe violations of multivariate normali-ty.Additionally,MLR procedures adjust/correct standard errors and the statistical significance of the unstandardized path coefficients,tak-ing into account multivariate non-normality.All other assumptions were met.
Multi-group structural path analysis was performed to evaluate the invariance of path coefficients among the distance education and tradi-tional education groups using EQS6.1(Bentler,2005).First,a fully-constrained,fully-saturated baseline model was established for both groups to examine the feasibility of the hypothesized path model pre-sented in Fig.1by specifying the12direct paths and thre
e error covari-ances(anxiety,frustration;anxiety,hope;and frustration,hope)and by imposing equality constraints on all direct paths and covariances.Sub-sequently exploratory model trimming(Wald test for dropping param-eters)and model building(Lagrange Multiplier[LM]test for adding parameters)procedures were interpreted in an effort to improve over-all modelfit of the baseline model.Next,equality constraints were indi-vidually removed for each ,freely estimated)that reached statistical significance at the p b.05level using the multivariate LMχ2univariate increment test for releasing equality constraints.This procedure was repeated until no further parameters'LMχ2univariate increment reached statistical significance.This model was then deemed thefinal model.Releasing equality constraints for any given parameter indicates that the parameter in question differs statistically significantly across the distance education and traditional education groups.Finally, theΔS-Bχ2(scaled chi squared difference)test was conducted to com-pare the ,fully-constrained,fully-saturated)model and thefinal ,released equality constraints).
5.Results
The Pearson product–moment correlation matrix presented in Table2demonstrates that most of the correlations among the vari-ables were moderate.
5.1.Equality among groups
Before proceeding with the multigroup path analyses,a series of analyses were conducted,including one independent samples t-test (for age as a continuous variable)and three binary logistic regres-sions to ascertain whether the two groups(distance education and traditional education)were homogenous in terms of pertinent demo-graphic variables—age,gender,previous experience with statistics, and previous experience with research methods.Course type served as the independent variable in all of the analyses,with each of the de-mographic variables serving as the dependent variable respectively. The results were all statistically non-significant,age,t(181)=0.13, p=.90,gender,χ2(1)=.02,p=.86,previous experience with statis-tics,χ2(1)=2.47,p=.11,and previous experience with research methods,χ2(1)=.08,p=.77.Therefore,because the groups did not significantly differ with respect to these demographic variables, they were not included as controls in the subsequent path analyses.
5.2.Path models
5.2.1.Baseline model
The baseline model for both groups with equality constraints im-posed on all path coefficients and co
variances specified in Fig.1demon-strated adequatefit to the data,S-Bχ2(24,N=291)=48.2671,p b.05, *NNFI=.92,*IFI=.96,*CFI=.96;however,the residual indices were relatively large,standardized root mean square residual(SRMR)=.10, *RMSEA=.06,90%CI[.03,.08].None of the respecifications suggested by the Wald test and LM test made theoretical sense based on the researchers'knowledge of the theory and variables under study. Hence,no respecifications were made to the baseline model.
5.2.2.Final model
Thefinal model with all statistically significant equality con-straints releasedfit the data well,S-Bχ2(18,N=291)=16.5712, p=.55,*NNFI=1.00,*IFI=1.00,*CFI=1.00,and exhibited low re-siduals,SRMR=.05,*RMSEA=.00,90%CI[.00,.05].The correlations between the exogenous variables(utility value,self-efficacy,and rel-evance)as well as the error correlations between all of the emotions variables were within normal range and statistically significant(see Table3).However,none of these relationships differed significantly among the groups.
5.2.3.Test of differences among nested models
As is evident from Table4,the S-B ScaledΔχ2test between the fully-constrained baseline model and t
hefinal model is statistically significant at the p b.001level of significance.Therefore,one can con-clude that thefinal model is a significant improvement in terms offit when compared to the fully-constrained baseline model.Further-more,this significant difference between the models indicates that those equality constraints which were released are statistically signif-icantly different between the distance education and traditional edu-cation groups.These differences among the two groups are reviewed and interpreted next.
5.3.Differences in path coefficients among the distance education and traditional education groups
Fig.2contains thefinal model with all path coefficient estimates and explained variances(R2)included.The R2values for the endoge-nous variables were moderate to high,ranging from.20to.51.The path coefficient from extrinsic utility value to anxiety was significant-ly different among the two groups.Whereas this path coefficient was non-significant and negative for the distance education group,it was significant and positive for the traditional education group.Moreover, the path from utility value to hope was significant for the distance ed-ucation group but non-significant for the traditional education group.
A reverse pattern was found in the path from relevance to hope, which was significant for the traditio
nal group but non-significant for the distance education group.Interestingly,the path from self-efficacy to hope,while significant for both groups,was significantly stronger in the traditional education group than the distance educa-tion group.Thefinal two significantly different path coefficients among the groups involved paths from emotions to learning strate-gies.The paths from frustration to learning strategies and anxiety to learning strategies were significant for the traditional education group but non-significant for the distance education group.The Table3
Correlations among exogenous variables and among emotions error terms. Parameter Distance a Traditional b
Pearson's r Pearson's r
Exogenous
UV,SE.33.33
UV,REL.59.54
REL,SE.30.39
Error
H,F−.51−.43
H,A−.51−.46
A,F.49.52
Key:UV=extrinsic utility value;SE=self-efficacy;REL=relevance;H=hope;
F=frustration;A=anxiety.
Note.All correlations were statistically significant at p b.05.
a N=72.
b N=219.
155
G.C.Marchand,A.P.Gutierrez/Internet and Higher Education15(2012)150–160
版权声明：本站内容均来自互联网，仅供演示用，请勿用于商业和其他非法用途。如果侵犯了您的权益请与我们联系QQ:729038198，我们将在24小时内删除。
发表评论