You are on page 1of 6

Factors for Success in Online CS1

Jennifer Campbell Diane Horton Michelle Craig


Dept of Computer Science Dept of Computer Science Dept of Computer Science
University of Toronto University of Toronto University of Toronto
campbell@cs.toronto.edu dianeh@cs.toronto.edu mcraig@cs.toronto.edu

ABSTRACT Like others [5, 16], we found that online students were
Enrollment in post-secondary online courses has been in- less likely to complete the course than face-to-face students.
creasing, but several studies have found that the drop rates However, for the students who did write the final exam,
in online courses are higher than in face-to-face. In our pre- exam scores in online and flipped were not significantly dif-
vious study comparing an online section of CS1 with a face- ferent. It is difficult to determine whether the poorer com-
to-face flipped section, we also found the drop rate higher in pletion rate in the online section is due to differences in
the online section. Given that we plan to continue offering populations or to the course format. A number of studies
online options for our students, we aim to identify factors as- have considered the factors leading students to drop out of
sociated with success in online CS1. In this paper, we exam- online courses in multiple disciplines. For a survey of papers
ine factors that are under students’ own control such as how from 1999 to 2009, see [15].
fully they participate in ungraded but important learning In this paper, we explore possible success factors in both
activities, and other factors that we may be able to manip- online and flipped offerings of CS1. There have been many
ulate and improve, such as students’ skills for self-regulated previous studies of success in traditional, lecture-based CS1
learning, and their sense of community in the course. We courses, including explorations of demographics, study be-
found important differences between the online and flipped haviours, previous academic experience or other computer
sections regarding what behaviours and attributes were asso- experience, cognitive skills, and student beliefs. See [2] for a
ciated with success. While completion of unmarked practice comprehensive survey of the pre-2006 work. Here we high-
exercises was a factor for both sections, test anxiety and self- light results relevant to the factors we explore in this paper.
efficacy were factors only for the online section, and intrinsic Rountree et al. [20] tested for associations between course
goal orientation was a factor only for the flipped section. outcomes and student demographics, background, and ex-
pectations. They found that desire to learn programming
was the strongest influence on passing the course, and that
Categories and Subject Descriptors grade expectations were a factor in the final grades. In a
K.3.2 [Computers and Education]: Computer and Infor- multi-institution study, Simon et al. [22, 8] reported that
mation Science Education—Computer Science Education final course grades were more strongly correlated with stu-
dents’ shallow or deep approach to learning than with prior
Keywords experience. Wilson and Shrock [24] found that the most im-
portant factors for success were students’ comfort level in
online; CS1; novice programming; self-regulated learning the course, math background, and attributing performance
on the midterm to luck (which contributed negatively). Ven-
1. INTRODUCTION tura [23] also considered comfort level, among other factors,
Enrollment in online courses is increasing and, as of 2012, and found that both comfort level and time spent in the lab
approximately 32% of post-secondary students in the United had more predictive value than cognitive and academic fac-
States took at least one online course [1]. In [12], we com- tors such as critical thinking and SAT scores. In contrast,
pare a CS1 course offered in two formats: online and face-to- Chinn et al. [6] did not find that study time was a factor in
face flipped. We found that the population of students who their traditional CS1, but that lecture attendance and prior
choose each section is different, with the online section at- experience both positively correlated with final grades.
tracting fewer CS majors, fewer first-year students and more Another factor for student success is self-regulated learn-
beginners, retakers, and part-time students. ing (SRL). “Self-regulated learning and performance refers
to the process whereby learners personally activate and sus-
Permission to make digital or hard copies of all or part of this work for personal or
classroom use is granted without fee provided that copies are not made or distributed
tain cognitions, affects, and behaviours that are systemati-
for profit or commercial advantage and that copies bear this notice and the full citation cally oriented toward the attainment of personal goals” [25,
on the first page. Copyrights for components of this work owned by others than the p. 1]. SRL has been studied extensively, and is an im-
author(s) must be honored. Abstracting with credit is permitted. To copy otherwise, or portant source of differences in achievement between stu-
republish, to post on servers or to redistribute to lists, requires prior specific permission
and/or a fee. Request permissions from permissions@acm.org. dents. Furthermore, helping students to strengthen their
ITiCSE ’16, July 09 - 13, 2016, Arequipa, Peru self-regulatory processes can improve student achievement [25].
c 2016 Copyright held by the owner/author(s). Publication rights licensed to ACM. Pintrich and de Groot [17] note that definitions of SRL vary,
ISBN 978-1-4503-4231-5/16/07. . . $15.00 but that three components are particularly important for
DOI: http://dx.doi.org/10.1145/2899415.2899457

320
classroom performance: cognitive strategies; metacognitive term that included the MSLQ [17]1 and a survey in the final
strategies such as monitoring cognition; and skills related to week that asked students to report on behaviours of interest.
the management of effort, such as persistence. They empha- We were also interested in how students’ sense of community
size that SRL skills are not sufficient for performance. Stu- might differ in the online section, so in the final survey we
dents must also possess the motivation to use those skills. included questions about interaction with each other and the
They developed and validated a now widely-used instrument instructor, as well as the Classroom scale of the Classroom
to assess SRL skills and motivational belief, the Motivated and School Community Inventory (CSCI) [21].
Strategies for Learning Questionnaire (MSLQ) [17, 18]. 855 students (118 online and 737 flipped) gave informed
In an online gerontology course, Cho and Shen [7] used consent and completed at least one survey. Regardless of
the MSLQ to study SRL motivations and behaviours. They whether they consented to participate in our study, students
found students’ intrinsic goal orientation and academic self- who completed both the initial and final survey earned 1%
efficacy to be important for academic achievements. towards their course grade.
There has been little work on self-regulated learning in We set a p-value of .05 for statistical significance. We
computer science. Falkner et al. [10] identify SRL strategies report exact values for p, to three decimal places, unless it
used by students in an introductory software development is < .001.
course, but they do not examine the relationship between
SRL and course outcomes. Consistent with the literature 3. RESULTS AND DISCUSSION
on SRL, Bergin et al. [3] found that greater use of metacog-
nitive and resource management strategies was linked to 3.1 Doing the unmarked exercises
better performance in an introductory programming course.
They also found that high intrinsic motivation and high task Flipped students spent most of their lecture time solving
value are linked to better performance and to greater use of unmarked exercises and online students worked on analo-
metacognitive and resource management strategies. Inter- gous exercises online. The exercises did not contribute to
estingly, they did not find the analogous relationships held course grades for either group. Nevertheless, we consider
for cognitive strategies. these exercises to be important for learning, and want stu-
dents to choose to do them. To measure whether students
did, we asked online students on the final survey “How of-
2. METHODOLOGY ten did you complete the ‘Rehearse’ (not-for-credit) work
on [the online system]?” For the flipped section, because the
2.1 The Course instructors did not record students’ exercise work, we use
This study was conducted at the University of Toronto, lecture attendance as an approximation of participation in
a research-intensive North American university. We have the in-class exercises. On the final survey we asked “Regard-
been offering our CS1 in a face-to-face, flipped format since less of whether you were enrolled in the online or on-campus
2013 [4, 14, 13]. In [12], we described our first offering of an sections, how often did you attend lecture?”2
online CS1 in Fall 2014 and compared it to our flipped CS1. Figure 1 shows online students’ self-reported completion
Our online CS1 was modelled after the flipped CS1, with of the Rehearse Exercises and flipped students’ self-reported
most face-to-face components replaced by online alterna- attendance at lectures.
tives. For all students, flipped and online, course content
was delivered through videos. The flipped students met
face-to-face during lecture periods to complete exercises as
described in [4], while the online students did analogous ex-
ercises asynchronously and online. In addition, to parallel
the just-in-time teaching that occurred in the flipped lec-
tures, the online students had access to explanatory videos
in which an instructor solved the exercises and addressed
anticipated misconceptions.
The online students completed their midterm online, while
the flipped students wrote a traditional midterm on paper.
The online midterm was limited to the question types avail-
able on our Learning Management System (Blackboard) and
by the fact that online students could not be supervised dur-
ing the test. Because of this, the midterm was worth only Figure 1: Self-reported Rehearse Exercise comple-
5% of the course grade, compared to 14% for the flipped tion for online students and lecture attendance for
students. All students wrote the same final exam in-person flipped students
and on paper, but the exam was worth 59% for online stu-
dents and 50% for flipped students. The rest of the course
work was the same for all students and was comprised of We hoped that online student participation in the un-
out-of-class weekly exercises and three larger assignments. marked Rehearse Exercises would be as good as attendance
1
2.2 The Study We used the 1990 version of the MSLQ. In the 1991 ver-
sion [18] of the MSLQ, used by Bergin et al. [3], many ques-
We wanted to determine how course outcomes were in- tions assume that the survey is conducted once the course
fluenced by students’ pre-course motivation and skills for is well underway, rendering it unsuitable for our study.
self-regulated learning, and by their behaviour during the 2
Because many online students were on campus and had
course. We therefore conducted a survey in the first week of access to lectures, we asked this question of both groups.

321
at lecture in the flipped section. This was not the case. The may have reported that they did not “complete” the exer-
mean response for completion of Rehearse Exercises in the cise while flipped students only had to report on their lecture
online section was 3.2 while the mean response for atten- attendance. Finally, both completion of Rehearse Exercises
dance in the flipped section was 4.1. A Welch two sample and attendance at lecture were self-reported.
t-test finds this difference significant (p < .001).
To test our assumption that participating in the exercises 3.2 Time spent on the course
is important for learning, we tested whether exercise partici- In the end-of-term survey we asked “How much time did
pation correlated with exam grades. For online students, we you spend on this course compared to other university courses
did not find a significant correlation between exam grades you have taken?” This question addresses the amount of em-
and self-reported completion of Rehearse Exercises. Since phasis students placed on the course, rather than the abso-
prior programming experience is known to be a strong pre- lute amount of time they invested.
dictor of outcomes in CS1 [11], we performed a linear re- In section 3.1, we saw that the completion rate of the
gression of exam grades modelled by prior experience3 and optional Rehearse Exercises for online students was signifi-
completion of Rehearse Exercises as shown in Table 1. Once cantly lower than the lecture attendance of flipped students.
prior programming experience was factored out, completion However, students in online and flipped gave similar re-
of Rehearse Exercises was a statistically significant factor sponses to our question about time. (Figure 2). The means
for exam grades. were 3.2 for online and 3.4 for flipped.

Estimate Std. Error t value Pr(>|t|)


(Intercept) 32.2438 4.6398 6.95 0.0000
prior.exp 4.2878 1.5431 2.78 0.0071
rehearse 2.1966 0.9957 2.21 0.0308

Table 1: Statistical models: effect of prior experi-


ence and completion of Rehearse Exercises on exam
grades for online students

The results are analogous for flipped students. There


was no significant correlation between exam grades and self-
reported lecture attendance, but once prior experience was
factored out, lecture attendance was significant. Table 2
shows the result of a linear regression of exam grades mod- Figure 2: Time Spent vs. Other Courses
elled by prior experience and lecture attendance.

Estimate Std. Error t value Pr(>|t|) We tested for a correlation between exam grades and rela-
(Intercept) 34.8040 2.3401 14.87 0.0000 tive time spent for students in each section. In related work,
prior.exp 3.7140 0.4913 7.56 0.0000 Chinn et al. [6] did not find a correlation between study time
attend 1.7146 0.5041 3.40 0.0007 and final results in a face-to-face CS1. For online students,
we found a weak negative correlation (Pearson’s product-
Table 2: Statistical models: effect of prior expe- moment correlation p = .015, n = 70, r = −.29). For
rience and lecture attendance on exam grades for flipped, there was no significant correlation between relative
flipped students time spent and exam grades. Because our question asked
about time spent relative to other courses rather than in
To summarize, for both online and flipped students, par- absolute terms, as asked by Chinn et al., the results cannot
ticipation in the not-for-credit exercises is a factor for final- be directly compared.
exam performance once prior experience is factored out. The negative correlation for online students between exam
Students who participate more in the optional activities do grades and relative time spent is surprising. To explore it,
better. We have no control over prior experience, but we can we examined the group of eight online students who reported
perhaps influence the students’ participation rates once they spending “much less” time on the course than on their other
enroll in the course. These results are consistent with the courses. Their mean exam score was 81.2%, much higher
study by Chinn et al. [6], which found that prior experience than the 65.6% exam average for all online students. Their
and lecture attendance correlated with final grades. mean prior experience score was 2.1 (versus 1.6 for all on-
There are several threats to the validity of these conclu- line students), which may explain why these students could
sions. Although the instructors report that flipped students spend relatively less time on the course and still do well.
who attended class participated well in the work, attendance
may not be a good approximation of work on the exercises. 3.3 Sense of community
Additionally, an online student who answered all the ques- The online students were required to be physically present
tions for a Rehearse Exercise, but did not get them correct, with each other and their instructor only once during the
3 course, at the final exam. Despite this, the instructors hoped
On the initial survey, we asked: “Many CSC108 students
have no prior programming experience, while others have to create as strong a sense of community in the online section
some. Rate the amount of prior programming experience as the flipped section, by fostering substantial interaction in
that you have.” on a scale from 1 (no experience) to 5 (a lot the online section, both among students, and between stu-
of experience). dents and their instructor. Students in both the online and

322
Online F2F
Question Average Average p-value
How much did you interact (both online and face-to-face) with your instructor? 1.942 2.589 < .001∗
How satisfied were you with your sense of connection to your instructor? 3.371 3.543 = .132
How much did you interact (both online and face-to-face) with your classmates? 2.125 3.280 < .001∗
How satisfied were you with your sense of community with your classmates? 3.124 3.274 = .227
CSCI Inventory score (out of a possible 50) 24.22 24.21 = .995

Table 3: Average response to questions about interaction and sense of community, and CSCI Inventory score,
by section. p-values are for a Welch two-sample t-test and an asterisk indicates statistical significance.

flipped sections were invited to participate in a shared on- self-efficacy scores were positively correlated with final exam
line discussion forum and over 900 (of 1100) students joined. scores (p = .009, n = 71, r = .30). Not only did students
Students were extremely active in the forum, posting over with a weak sense of self-efficacy do more poorly on the
3600 times, and each reading, on average 141 posts. The in- exam, but they were also less likely to finish the course: the
structors were also very active in the forum; they attempted mean self-efficacy scores of those who dropped the course
to read all student posts, and themselves made over 950 was 4.5 compared to 5.0 for those who completed the course
posts. The TAs made an additional 1500 posts. In addition (passed or failed). A Welch two sample t-test finds this dif-
to the face-to-face office hours available to all students, the ference significant (p = .029). Neither of these relationships
online instructor and a TA also held online office hours that held for the flipped students.
were available to all. Unfortunately, the online office hours Our results suggest that one might improve outcomes for
were poorly attended; the requirement to install additional online students if we can boost their self-efficacy. Cho and
software may have been a barrier. Shen [7] emphasize that having a strong teaching presence
To determine whether these efforts were successful in cre- contributes to students’ sense of self-efficacy, as does encour-
ating an equally strong sense of community, we asked four aging students to set goals that are challenging but achiev-
questions in the final survey about sense of community, and able so that they can experience success.
also administered the Classroom scale of the CSCI. Table 3
shows the results. Online students reported significantly less Intrinsic value
interaction with their instructor and their classmates, yet Students in the online and flipped sections also had nearly
their sense of connection to their instructor and classmates identical scores for intrinsic value (mean 5.4 online; 5.5 flipped).
was not significantly different, and their CSCI scores were There was no correlation between exam grade and intrinsic-
nearly identical (mean 24.4 online; 24.2 flipped). value score for the online or flipped sections. However, we
We wondered whether students with a stronger sense of found that once prior programming experience is controlled
community would do better in the course, so we examined for, intrinsic value is a statistically significant factor for exam
the correlation between CSCI scores and exam grades. We grades, but only for the flipped section (see Table 4).
found that this correlation was not statistically significant
for either the online or flipped sections. Since the CSCI Estimate Std. Error t value Pr(>|t|)
scale was administered at the end of the course, students (Intercept) 34.8479 3.7282 9.35 0.0000
who dropped earlier because they were dissatisfied with their prior.exp 3.1880 0.5060 6.30 0.0000
sense of community would not be represented in these re- intrinsic.val 1.3976 0.6758 2.07 0.0391
sults. To determine whether sense of community was re-
lated to successful completion of the course, if not with exam Table 4: Statistical models: effect of prior experi-
grades, we examined the mean CSCI scores of students who ence and intrinsic value on exam grades for flipped
passed and who failed the course; a Welch two-sample t-test students
showed that they were not significantly different. Our results
are consistent with Drouin’s findings that neither achieve- Bergin et al. [3] found that, in their face-to-face course,
ment nor retention is related to sense of community [9]. students with high intrinsic motivation performed better in
programming than those with low intrinsic motivation. This
3.4 Motivational beliefs is consistent with our result for the flipped section. Other
studies of online and face-to-face courses, have also shown
The MSLQ [17] includes three scales about motivation:
that intrinsic motivation is positively related, directly or in-
self-efficacy (the sense that one is able to perform a task),
directly, to course outcomes [3, 7, 19].
intrinsic value (being motivated by goals such as a desire for
Although we did not find a significant correlation between
mastery, rather than by extrinsic goals such as grades) and
intrinsic motivation and exam scores for the online section,
test anxiety. For each, we compared scores between sections,
we hope that boosting intrinsic motivation may still help im-
and looked for evidence of a relationship to course outcomes.
prove outcomes in not only flipped, but also online. Rakes
and Dunn [19] survey the literature on increasing intrin-
Self-efficacy sic motivation. They identify factors that they believe are
Online and flipped students had nearly identical mean scores particularly relevant to online courses, including creating a
(rounding to 4.9 for both groups) for self-efficacy. However, sense of community, projecting a supportive instructional
we found evidence of a relationship between self-efficacy and style, encouraging a perception of competence, presenting
outcomes only in the online section. For online students, challenges, and encouraging autonomy. In [7], they suggest

323
introducing problem-based learning (PBL) and framing stu- of the MSLQ. However, as with the cognitive strategy scale,
dents’ learning in terms of intrinsic goals. both versions of the MSLQ emphasize reading and studying,
and neglect practise of skills.
Test anxiety
Scores on the test-anxiety scale were not significantly dif- 4. CONCLUSIONS
ferent for students in the two sections (mean 3.6 online; 3.8
flipped). We found a negative correlation between test anxi- Having observed a higher drop rate in our online CS1 than
ety and exam grade for the online section (p < .001, n = 79, in the flipped section, we aimed to identify students’ at-
r = −.38); students with greater test anxiety did worse on tributes and behaviours that may be factors for success in
the exam. Surprisingly, in the flipped section there was no CS1 generally and for these formats in particular. We have
significant correlation between test anxiety and exam score. identified several key differences between students in online
Test anxiety was measured when the MSLQ was adminis- and flipped. Because our online students choose to take the
tered at the beginning of the course. It may be that the in- course in the online format, these results may not generalize
class, on-paper exercises that the flipped students complete to a context where online is the only format available. Sev-
during the term combat their anxiety prior to the paper- eral of our results suggest actions an instructor can take to
based exam, either by reducing anxiety or by increasing stu- improve an online course.
dents’ test-taking skills sufficiently to counteract the effect For both the online and flipped sections, once prior experi-
of their anxiety. Online students complete their analogous ence is controlled for, completion of the unmarked exercises
exercises online, so they get less practice working on paper. is a statistically significant factor for exam scores. However,
the completion rate of Rehearse Exercises by online students
3.5 SRL Strategies was significantly lower than the flipped students’ rate of lec-
ture attendance. In future offerings, we will try to boost
Cognitive strategy use Rehearse Exercise completion, for example by incorporating
more reminders from the instructor and by following up with
Scores on the cognitive strategy scale were not significantly emails during the term to students who have not completed
different across sections (mean 4.9 online; 5.0 flipped). We the exercises.
hypothesized that there would be a positive correlation be- We found self-efficacy to be correlated with final exam
tween cognitive strategy and final exam scores. This hypoth- scores for online, but not for flipped. For online students, the
esis was not supported. The correlation was non-significant mean self-efficacy scores for those who dropped, passed, and
for the flipped section, and there was a weak negative cor- failed were also significantly different. Cho and Shen report
relation for the online section (Pearson’s product-moment that instructor presence is key for increasing self-efficacy [7],
correlation p = .025, n = 76, r = −.26). so we will aim to increase instructor-student interactions in
Bergin et al. [3] also failed to find a relationship between future offerings of the online course.
cognitive strategy and programming performance. They For our flipped students, once we control for prior pro-
suggest that the strategies measured by the MSLQ may not gramming experience, intrinsic value is a statistically signif-
be as useful in learning introductory object-oriented pro- icant factor for exam grades. Although we did not find the
gramming as in other academic areas. Indeed, the MSLQ same for online, we feel it is still worthwhile to try to boost
(both the 1991 version that they used and the 1990 version students’ intrinsic value in both online and flipped. Based
that we used) places substantial emphasis on skills related on the five factors [19] identified for encouraging intrinsic
to reading and studying, and little emphasis on mastering motivation, we plan to host informal get-togethers of the
a skill, a key component of any CS1 course. An instrument instructor and students, and include additional encouraging
that includes strategies related to mastering a skill, or to and supportive comments in communications with students.
learning to program specifically, may be of greater value in For online students, there was a negative correlation be-
predicting outcomes in CS1. Falkner’s work [10] to identify tween text anxiety and final exam scores, but there was no
discipline-specific cognitive and metacognitive skills used by significant correlation between the two for flipped students.
students learning to program is a contribution towards this. The difference may be explained by the fact that the exam
Self-regulation is written on paper, but only flipped students were asked
to regularly practice solving problems on paper. In future
The self-regulation scale assesses metacognitive strategies offerings, we plan to assign online students more unmarked
and management of effort. As for all the other MSLQ scales, exercises on paper in an effort to better prepare them for
scores were not significantly different between sections (mean the final exam. One possibility is to assign a question each
4.8 online; 4.9 flipped). We found no evidence that self- week from a past exam.
regulation, as measured by the MSLQ, is a factor for success Online education will undoubtedly continue to grow, but
for students in either section: for both the online and flipped much of what we know about teaching and learning comes
sections, the correlation between self-regulation and exam from experience and research in face-to-face courses. If we
score was not statistically significant, linear regression did are to provide effective learning environments for students
not reveal a significant relationship to exam score even when in an online context, we must continue to research best prac-
prior-experience was accounted for, and there was no signifi- tices and apply them in our teaching.
cant difference between the self-regulation scores of students
who dropped the course and students who completed it.
In contrast, Bergin et al. [3] did find that both stronger 5. ACKNOWLEDGMENTS
metacognitive skills and better management of resources The development of the online course was funded by the
were associated with higher grades. Our failure to replicate Ontario Ministry of Training, Colleges and Universities. We
this result may be due to differences between the versions thank Hedieh Najafi for assistance with our literature search.

324
6. REFERENCES D. Zingaro. Comparing outcomes in inverted and
[1] I. E. Allen and J. Seaman. Changing Course: Ten traditional cs1. In Proceedings of the 2014 Conference
Years of Tracking Online Education in the United on Innovation & Technology in Computer Science
States. ERIC, 2013. Education, ITiCSE ’14, pages 261–266, 2014.
[2] S. Bergin and R. Reilly. Predicting introductory [15] Y. Lee and J. Choi. A review of online course dropout
programming performance: A multi-institutional research: implications for practice and future research.
multivariate study. Computer Science Education, Educational Technology Research and Development,
16(4):303–323, 2006. 59(5):593–618, 2011.
[3] S. Bergin, R. Reilly, and D. Traynor. Examining the [16] Y. Levy. Comparing dropouts and persistence in
role of self-regulated learning on introductory e-learning courses. Computers & education,
programming performance. In Proceedings of the First 48(2):185–204, 2007.
International Workshop on Computing Education [17] P. R. Pintrich and E. V. De Groot. Motivational and
Research, ICER ’05, pages 81–86, 2005. self-regulated learning components of classroom
[4] J. Campbell, D. Horton, M. Craig, and P. Gries. academic performance. Journal of educational
Evaluating an inverted CS1. In Proceeding of the 45th psychology, 82(1), Mar. 1990.
ACM Technical Symposium on Computer Science [18] P. R. Pintrich, D. Smith, T. Garcia, and
Education, SIGCSE ’14, pages 307–312, 2014. W. McKeachie. A manual for the use of the motivated
[5] S. Carr. As distance education comes of age, the strategies for learning questionnaire. technical report
challenge is keeping the students. Chronicle of higher 91-b-004, 1991.
education, 46(23), 2000. [19] G. C. Rakes and K. E. Dunn. The impact of online
[6] D. Chinn, J. Sheard, A. Carbone, and M.-J. Laakso. graduate students’ motivation and self-regulation on
Study habits of cs1 students: What do they do outside academic procrastination. Journal of Interactive
the classroom? In Proceedings of the Twelfth Online Learning, 9(1):78–93, 2010.
Australasian Conference on Computing Education - [20] N. Rountree, J. Rountree, A. Robins, and R. Hannah.
Volume 103, ACE ’10, pages 53–62. Australian Interacting factors that predict success and failure in a
Computer Society, Inc., 2010. cs1 course. In Working Group Reports from ITiCSE
[7] M.-H. Cho and D. Shen. Self-regulation in online on Innovation and Technology in Computer Science
learning. Distance Education, 34(3):290–301, 2013. Education, ITiCSE-WGR ’04, pages 101–104, 2004.
[8] M. de Raadt, M. Hamilton, R. Lister, J. Tutti, Simon, [21] A. P. Rovai. The classroom and school community
K. Sutton, B. Baker, I. Box, Q. Cutts, J. Hamer, inventory: Development, refinement, and validation of
M. Petre, and D. Tolhurst. Approaches to learning in a self-report measure for educational research. The
computer programming students, and its effect on Internet and Higher Education, 7(4):263–280, 2004.
success. In Proceedings of Higher Education in a [22] Simon, S. Fincher, A. Robins, B. Baker, I. Box,
Changing World, Higher Education Research and Q. Cutts, M. de Raadt, P. Haden, J. Hamer,
Development Society of Australia Conference, pages M. Hamilton, R. Lister, M. Petre, K. Sutton,
407–414, 2005. D. Tolhurst, and J. Tutty. Predictors of success in a
[9] M. A. Drouin. The relationship between students’ first programming course. In Proceedings of the 8th
perceived sense of community and satisfaction, Australasian Conference on Computing Education -
achievement, and retention in an online course. Volume 52, ACE ’06, pages 189–196, Darlinghurst,
Quarterly Review of Distance Education, 9(3):267–284, Australia, Australia, 2006. Australian Computer
2008. Society, Inc.
[10] K. Falkner, R. Vivian, and N. J. Falkner. Identifying [23] P. R. Ventura. Identifying predictors of success for an
computer science self-regulated learning strategies. In objects-first cs1. Computer Science Education,
Proceedings of the 2014 Conference on Innovation & 15(3):223–243, 2005.
Technology in Computer Science Education, ITiCSE [24] B. C. Wilson and S. Shrock. Contributing to success
’14, pages 291–296, New York, NY, USA, 2014. ACM. in an introductory computer science course: A study
[11] D. Hagan and S. Markham. Does it help to have some of twelve factors. SIGCSE Bull., 33(1):184–188, Feb.
programming experience before beginning a 2001.
computing degree program? In Proceedings of the 5th [25] B. J. Zimmerman and D. H. Schunk. Self-regulated
Annual ITiCSE Conference on Innovation and learning and performance: An introduction and
Technology in Computer Science Education, ITiCSE overview. In B. J. Zimmerman and D. H. Schunk,
’00, pages 25–28, 2000. editors, Handbook of Self-Regulation of Learning and
[12] D. Horton, J. Campbell, and M. Craig. Online cs1: Performance, chapter 1, pages 1–12. Taylor and
Who enrols, why, and how do they do? In Proceeding Francis, New York, 2011.
of the 47th ACM Technical Symposium on Computer
Science Education, SIGCSE ’16, pages 323–328, 2016.
[13] D. Horton and M. Craig. Drop, fail, pass, continue:
Persistence in cs1 and beyond in traditional and
inverted delivery. In Proceedings of the 46th ACM
Technical Symposium on Computer Science Education,
SIGCSE ’15, pages 235–240, 2015.
[14] D. Horton, M. Craig, J. Campbell, P. Gries, and

325

You might also like