Professional Documents
Culture Documents
Chemistry
View Journal
Education Research
and Practice
Accepted Manuscript
This article can be cited before page numbers have been issued, to do this please use: B. Ferrell, M. M.
Phillips and J. Barbera, Chem. Educ. Res. Pract., 2016, DOI: 10.1039/C6RP00148C.
www.rsc.org/cerp
Page 1 of 24 Chemistry Education Research and Practice
View Article Online
DOI: 10.1039/C6RP00148C
1
2
3
Connecting achievement motivation to performance in general chemistry
4
5
6 Brent Ferrell1, Michael M. Phillips2, and Jack Barbera3
1
Department of Natural Sciences and Mathematics, Lee University, Cleveland, TN
1
2
3
properties of four scales that measured self-efficacy, personal interest, situational
4
5 interest, and effort beliefs (Ferrell and Barbera, 2015). In the present study, we
6 extend upon our previous work by investigating the connections between these
1
2
3
with low self-efficacy. Although a display of effort is not equivalent to believing
4
5 that effort will lead to positive outcomes, we argue that positive effort beliefs
6 toward a task likely precede the exertion of effort. The relationship between
1
2
3
between situational interest and course performance. Harackiewicz et al. (2000)
4
5 investigated the short and long-term effects of college students’ goal orientations,
6 interest, and performance in an introductory psychology class. The authors
1
2
3
academic performance in chemistry, one cannot ignore the underlying non-
4
5 cognitive beliefs and processes. Several non-cognitive factors such as self-
6 efficacy (Zusho et al., 2003), attitude (Xu et al., 2013), self-concept (Lewis et al.,
18 Our aim in this study was to explore the possible relationships that exist
19 between self-efficacy, effort beliefs, personal interest, and situational interest as
20 well as their connection to final grades in an introductory chemistry course.
21 Toward this end, we used multiple regression and path analysis to test for
22 plausible models that best represent these relationships. Previously, we adapted
23
24 and modified four scales intended to measure these motivation-related variables
25 in a chemistry setting (Ferrell and Barbera, 2015). We demonstrated that all four
26 scales produced sufficient evidence for validity and reliability with both initial and
27 cross-validation samples of first-semester general chemistry students, a
28 necessary precursor to any further studies (Arjoon et al., 2013; Barbera and
29
30
VandenPlas, 2011; Brandriet et al., 2013). Our data collection and subsequent
31 analyses for the present study were guided by the following two research
32 questions:
33 (1) What are the connections among self-efficacy, interest, and effort beliefs
34 with general chemistry students?
35 (2) To what extent do self-efficacy, interest, and/or effort beliefs predict
36
37 course performance in general chemistry?
38 We hypothesized that self-efficacy would be the best predictor of course
39 performance in line with what others have reported (Bong, 2001; Pajares and
40 Miller, 1995). In addition, we hypothesized that personal interest would predict
41 situational interest, a relationship that has been theoretically established and
42
43
tested in several studies (Harackiewicz et al., 2008; Hidi and Renninger, 2006;
44 Nieswandt, 2007), and these would be related to course performance. Based on
45 the lack of research on effort beliefs in college settings, we did not have specific
46 hypotheses regarding these factors. We believe the results from this study will
47 provide valuable insight for instructors interested in measuring the motivational
48
climate of their classrooms as well as those who want to expand the evidence of
49
50 impacts related to curriculum changes.
51
52 Methods
53 Participants
54 Participants for this study were recruited from three sections of a first-semester
55
56
general chemistry course at a mid-sized Rocky Mountain region university during
57
58
59
60
Chemistry Education Research and Practice Page 6 of 24
View Article Online
DOI: 10.1039/C6RP00148C
1
2
3
the fall of 2014. This is a required course for several science and health-related
4
5 majors, and is the first in a two-semester sequence of general chemistry.
6
18 aspects for the courses. In addition to using the same textbook and homework
19 system, the instructors agreed on the timing of assessments and weight
20 percentages of assignments. Each course had an equal number of online
21 homework assignments, weekly quizzes, and hour exams. While the content
22 coverage of each was similar, based on the pacing of each instructor, they did
23
24 not use matched homework or assessment items for security reasons. Both
25 instructors administered an American Chemical Society examination for their final
26 exam. Lecture was the main teaching style for all courses. While both instructors
27 supplemented lectures with brief activities, problem solving tasks, and virtual or
28 physical demonstrations, neither would categorize their style as active learning.
29
30
31 Measures
32 Interest Scales. To assess interest at two time points during the semester,
33 personal and situational interest scales were needed. Both scales were originally
34 developed by Harackiewicz et al. (2008) and very similar items were used in a
35 subsequent study by Linnenbrink et al. (2010). We adapted the items from both
36
37 scales for use in college-level chemistry and found the psychometric properties
38 of each to be acceptable (Ferrell and Barbera, 2015). The personal interest scale
39 is comprised of seven items, three of which are feeling-related items, and the
40 remaining four are value-related. The situational interest scale is comprised of
41 eight items, with an equal split of feeling-related and value-related items. The
42
43
personal interest scale is designed to measure the interest that students already
44 have towards chemistry prior to taking the course. The situational interest scale,
45 on the other hand, is comprised of items that assess interest aimed at students’
46 feelings and value toward the material covered at the course level (Linnenbrink-
47 Garcia et al., 2013). Both scales are measured using a five-point Likert scale
48
ranging from “strongly disagree” to “strongly agree”.
49
50
51 Self-efficacy. To measure self-efficacy, select items from the College Chemistry
52 Self-Efficacy Scale were used (CCSS; Uzuntiryaki and Aydin, 2009), all of which
53 were previously tested for adequate psychometric properties. We originally
54 tested a scale that consisted of eight items, two of which were removed due to
55
56
poor model fit (Ferrell and Barbera, 2015). The scale used in the present study
57 consists of six items that ask students to rate how well they could complete tasks
58
59
60
Page 7 of 24 Chemistry Education Research and Practice
View Article Online
DOI: 10.1039/C6RP00148C
1
2
3
that would generally be encountered in an introductory chemistry course. The
4
5 items are measured on a five-point Likert-type scale ranging from “very poorly” to
6 “very well”.
18
19 Academic Performance. Although there are many ways to measure academic
20 performance in a college classroom, course grade is the most common. We
21 chose to use course grade percentage because it represents the final outcome of
22 the course and is the highest stake for students. Final course grade percentages
23
24 were based on content measures only including homework, quizzes, mid-term
25 exams, and the ACS final exam. No points were awarded for participation or
26 attendance, and the co-requisite lab was a separate grade entirely. Students with
27 incomplete grades were not included in the analyses.
28
29
30
Data Collection
31 Data for this study were collected at two time points during the lecture period of
32 each section. Time 1 (T1) data were collected during the first week of the
33 semester and time 2 (T2) data were collected during the 12th week of the
34 semester. On the days when data were collected, the researcher made an
35 announcement informing students that their participation was voluntary and that
36
37 their identities would be kept completely confidential. The students were then
38 provided the survey packet along with a consent form that was approved by the
39 Institutional Review Board. Students were given approximately 10 minutes to
40 complete all of the scales (19 items at time 1 and 20 items at time 2) and seven
41 demographic items (only given at time 1). Due to the nature of this study, only
42
43
students who completed all items at both time points were included in
44 subsequent analyses.
45
46 Data Analysis
47
48
Descriptives. Descriptive statistics including mean, standard deviation, skew
49
50 and kurtosis values were determined for all scales at both time points. We
51 considered acceptable skew and kurtosis values to be -1 to +1 (Huck, 2012).
52 Statistical Package for the Social Sciences (SPSS) 20.0 software was used for
53 these analyses.
54
55
56
Multiple Regression. Multiple regression analysis is a method used to predict or
57 explain the variation in a dependent variable by examining its relations to several
58
59
60
Chemistry Education Research and Practice Page 8 of 24
View Article Online
DOI: 10.1039/C6RP00148C
1
2
3
independent variables (Pedhazur, 1997). By utilizing multiple regression
4
5 analyses, the researcher is able to identify and separate the individual effects of
6 distinct independent variables on a dependent variable. Multiple regression
1
2
3
generate our path diagrams, a set of a priori path models were tested using
4
5 LISREL 8.80 (Jöreskog and Sörbom, 2006) with the Maximum Likelihood (ML)
6 estimator.
18 students (79.3%) were at the freshman or sophomore level and 81.7% reported
19 this class as being their first chemistry class in college. With regard to
20 race/ethnicity, 73.5% were Caucasian, 17.1% were Hispanic, 4.1% were African
21 American, 1.2% were Asian American, and 4.1% were self-identified as Other.
22 The breakdown of majors was as follows: 14% chemistry, 23% other science
23
24 (biology, physics, mathematics), and 63% were non-science (nursing, sports and
25 exercise science, other).
26
27 Data Screening and Descriptive Statistics. All data sets were screened for
28 careless responses (e.g., selection of the same option for every item) and none
29
30
were found. Descriptive statistics and internal consistency estimates (Cronbach’s
31 alpha) for the students’ scores are reported in Table 1. A final combined sample
32 size of 170 was analyzed from all three sections of the course. The skew and
33 kurtosis values were within the range of -1 to +1, with the exception of the value-
34 related personal interest scale. As a result, the inferences drawn from the scores
35 from this scale should be interpreted with caution as multiple regression and
36
37 maximum likelihood estimation are not robust to univariate nonnormality (Curran
38 et al., 1996; Osborne and Waters, 2002). The internal consistency analysis for
39 each scale reveals that Cronbach’s alpha (α) values are acceptable (> 0.70), with
40 the exception of the effort beliefs scores at T1 (α = 0.68). This borderline internal
41 consistency value was taken into consideration when interpreting the results
42
43
below. The descriptive statistics reported here are comparable to those reported
44 in our previous study using these scales, suggesting consistency in the
45 measurement (Ferrell and Barbera, 2015).
46
47 Table 1. Descriptive statistics for each measured scale (n = 170)
48
49 Scale (Time) Mean SD Skewness Kurtosis Cronbach’s α
50 PI – feel (T1) 3.25 0.90 -0.30 -0.20 0.92
51
52 PI – value (T1) 3.87 1.00 -1.61 2.68 0.89
53 EB (T1) 4.06 0.52 -0.26 -0.54 0.68
54
SE (T1) 2.90 0.71 -0.28 0.19 0.88
55
56 SI – feel (T2) 3.25 0.97 -0.39 -0.30 0.94
57 SI – value (T2) 3.51 0.84 -0.36 -0.03 0.86
58
59
60
Chemistry Education Research and Practice Page 10 of 24
View Article Online
DOI: 10.1039/C6RP00148C
1
2
3
EB (T2) 3.94 0.68 -0.75 0.43 0.83
4
5 SE (T2) 3.72 0.60 -0.50 0.36 0.85
6
Final % 75.81 12.68 0.25 -0.54 --
18
19
20
Table 2. Pearson correlations of all scales and final course grade
21 Scale PI-feel PI-value EB T1 SE T1 SI-feel SI-value EB T2 SE T2 Fpct
22 PI-feel -
23 PI-value 0.668* -
24 EB T1 0.409* 0.589* -
25
SE T1 0.319* 0.184* 0.059 -
26
27 SI-feel 0.503* 0.197* 0.214* 0.243* -
28 SI-value 0.361* 0.188* 0.185* 0.195* 0.662* -
29 EB T2 0.177* 0.111 0.423* 0.075 0.568* 0.486* -
30
SE T2 0.322* 0.218* 0.298* 0.400* 0.541* 0.474* 0.495* -
31
32 Fpct 0.151* 0.120 0.102 0.228* 0.392* 0.231* 0.312* 0.536* -
33 Note: PI – Personal interest, EB – Effort beliefs, SE – Self-efficacy, SI – Situational interest, Fpct – Final course
percent
34
*Significant at p < 0.05.
35
36
37 Final course grade was regressed on self-efficacy, effort beliefs, and both
38 feeling and value-related components for interest (personal for T1 and situational
39 for T2) at both time points. We used a hierarchical regression method to test the
40 incremental effects of T2 variables over the T1 variables in the model. The first
41
model was significant (F(4,165) = 2.741, p = 0.030, R2 = 0.062, R2 adjusted =
42
43 0.040) with only one significant main effect, self-efficacy T1 (β = 0.205, p =
44 0.011). The second model was significant as well (F(8,161) = 10.074, p < 0.001,
45 R2 = 0.334, R2 adjusted = 0.300, R2 change = 0.271, p < 0.001). Only situational
46 interest feel (β = 0.277, p = 0.011) and self-efficacy T2 (β = 0.477, p < 0.001)
47 were significant main effects in the second model. None of the T1 variables
48
49
accounted for a significant amount of variance in final course grade when the T2
50 variables were introduced into the model. This data suggests that variables
51 measured later in the semester account for significantly more variance in final
52 course grade than T1 variables, even after the variance from T1 variables were
53 accounted for.
54
55
56 Path Analyses. Path analysis was used to test a set of path models, which
57 describe a network of relationships simultaneously. Path analysis extends
58
59
60
Page 11 of 24 Chemistry Education Research and Practice
View Article Online
DOI: 10.1039/C6RP00148C
1
2
3
beyond multiple regression by allowing for relationships among predictor
4
5 variables to be tested as well as allowing for defined constraints on the model. As
6 with the multiple regression, all models were tested with the entire data set (n =
18 in model 1.
19 The results showed that model 1 was not a good fit to the data. Only the
20 GFI (0.91) was within an acceptable range (see Table 3). The R2 value for model
21 1 indicates that 25% of the variance in course grade can be explained by the way
22 students responded to the scales as outlined by the model.
23
24
25 Table 3. Fit statistics for each model tested, bold indicates value within
26 acceptable range for the type of statistic
27 Model 𝜒 ! (df) p-value RMSEA CFI GFI R
2*
28
29 1 44.25 (5) < 0.001 0.22 0.81 0.91 0.25
30 2 13.58 (4) < 0.01 0.12 0.96 0.97 0.28
31
3 10.46 (3) 0.02 0.12 0.97 0.98 0.29
32
33 4 3.59 (3) 0.31 0.03 0.99 0.99 0.30
34 *R reported as variance accounted for in final grade.
2
35
36
37 To explore the possibility of model misspecification, we examined the
38 modification indices (MIs). Modification indices are equivalent to a one degree of
39 freedom 𝜒 ! statistic that can only be applied to parameters which are fixed. The
40 MIs provide an estimate of how much the overall 𝜒 ! statistic will drop when the
41
parameter in question is freely estimated (Byrne, 2013). The MIs for model 1
42
43 indicated that correlating the disturbance terms of self-efficacy T2 and situational
44 interest feel would improve the overall fit of the model. Disturbance terms are
45 analogous to error terms in multiple regression and structural equation modeling.
46 They are estimates of the variance not accounted for by the incoming path or
47 paths to a given term. By allowing the correlations of the disturbance terms to be
48
49
freely estimated, we utilized a method known as correlated uniqueness (CU)
50 (Kenny, 1979). Although somewhat contentious in the literature, CU allows the
51 researcher to account for common method variance that can be particularly
52 problematic when using the same survey and scale type to measure multiple
53 variables (Marsh and Bailey, 1991; Podsakoff et al., 2003). When we re-specified
54
the model to include this change (model 2, Fig. 1), the fit improved dramatically
55
56 over model 1 (see Table 3). Nevertheless, the 𝜒 ! goodness-of-fit test was still
57 significant and the RMSEA value was not within an acceptable range.
58
59
60
Chemistry Education Research and Practice Page 12 of 24
View Article Online
DOI: 10.1039/C6RP00148C
1
2
3
4
5 0.32*# 0.32*#
6
1
2
3
related situational interest in chemistry. All other significant paths from model 2
4
5 were similar in direction and magnitude in model 3.
6 When a path was added from personal interest feel to self-efficacy T2 in
18
19 0.32*# 0.32*#
20
21
22 Personal# Self.Efficacy# Personal# Self.Efficacy#
23 Interest#Feel# T1# Interest#Feel# T1#
24
25 0.39*# 0.13# 0.40*# 0.50*# 0.23*# 0.29*#
26
27 Situa<onal# Self.Efficacy# Situa<onal# Self.Efficacy#
28 Interest#Feel# T2# Interest#Feel# T2#
29
30
31 0.80# 0.84# 0.75# 0.82#
32
d1# d2# d1# d2#
33
34
35 0.36*#
0.38*#
36
37
38
39 0.14# 0.46*# 0.15# 0.45*#
40
41 Final#Grade# Final#Grade#
42
43 0.71# 0.70#
44
45 d3# d3#
46
47 Model 3 Model 4
48
49
Figure 3. Path diagram of models 3 and 4. Standardized path coefficients are
50 reported. *Significant at p < 0.05
51
52 Discussion
53 In this study, we explored the connections between self-efficacy, personal
54 interest, situational interest, and effort beliefs as well as their impact on course
55
56 grade in a first-semester general chemistry course. Participants were evaluated
57 on these measures at two time points (week 1 and week 12). This work builds on
58
59
60
Chemistry Education Research and Practice Page 14 of 24
View Article Online
DOI: 10.1039/C6RP00148C
1
2
3
our prior study (Ferrell and Barbera, 2015) by investigating the temporal links and
4
5 interconnections, via path analysis, that exist between these motivational
6 variables. In addition, we analyzed the extent to which students’ grades were
18 we see that the strongest predictor of course grade was self-efficacy at T2 that
19 accounts for 20.3% of the total variance explained in the model.
20
21 Multiple Regression Analysis. To begin, we regressed final course grade on all
22 measured variables at both time points. Based on prior research, we expected
23
24 self-efficacy and interest to be potential predictors of course grade (Gore, 2006;
25 Harackiewicz et al., 2000; Smist, 1993; Zusho et al., 2003). Our results are
26 concurrent with this, suggesting that self-efficacy, measured at the end of the
27 semester, is the strongest predictor of course grade. Feeling-related situational
28 interest also accounts for a significant portion of the variance in course grade in
29
30
some of our models. These results were obtained even when personal interest
31 feel and T1 self-efficacy were accounted for in a hierarchical regression model.
32 Effort beliefs at either time point was found to not be a significant predictor of
33 course grade in our regression model. This is in line with what has been
34 previously reported (Blackwell et al., 2007; Jones et al., 2012). However, effort
35 beliefs T2 did exhibit strong correlations with all other T2 variables (see Table 2),
36
37 demonstrating a clear relationship with distinct, motivation-related constructs. To
38 further understand effort beliefs as a psychological construct, we are conducting
39 a qualitative investigation among general chemistry students that will be reported
40 at a later date.
41 It is important to note here that the timing of measurement should be a
42
43
central consideration when assessing students’ motivation. At the start of the
44 semester, self-efficacy T1 was a very weak predictor of final course grade and
45 personal interest was not a significant predictor. By the end of the semester, self-
46 efficacy T2 and situational interest feel accounted for a much larger portion of the
47 variance than the T1 variables. The predictive effect of self-efficacy measured in
48
the middle of the semester, versus the beginning, on performance has been
49
50 observed consistently by others (Bong, 2001; Gore, 2006; Lee et al., 2014). We
51 postulate this could be due to two reasons. First, students are not well calibrated
52 in their confidence when they walk in the door on the first day of chemistry.
53 Hence, they may under or over-estimate their capability to complete certain tasks
54 they will encounter in the course. Others have reported evidence of high school
55
56
students’ overconfidence in math, but suggested that college students would
57 likely have better calibration (Pajares and Kranzler, 1995). Although most of the
58
59
60
Page 15 of 24 Chemistry Education Research and Practice
View Article Online
DOI: 10.1039/C6RP00148C
1
2
3
students in this study had chemistry in high school, some did not, which would
4
5 further complicate the issue of calibration because of the lack of knowledge, skill,
6 and prior attainments (Gist and Mitchell, 1992; Pajares, 1996). By the end of the
1
2
3
those who do not. However, as Nauta et al. (2002) found, the directionality could
4
5 change with subsequent measurements of interest and self-efficacy. Thus, we do
6 not suggest our results are in conflict with what has been reported previously, but
1
2
3
more than half of the participants provided all of the necessary criteria for
4
5 inclusion in the final analyses. Finally, those who were included were primarily
6 female and Caucasian. Thus, the generalizability of our results to other
18 properties of each scale used in the present study, a critical first step to any
19 substantive investigation using self-report data. By having measures that have
20 been properly designed and target the variables appropriately is critical to
21 reducing the effects of systematic measurement error. Despite our efforts, one
22 reliability estimate is borderline (EB (T1), α = 0.68), therefore the statistical power
23
24 and parameter estimates in our model may be incorrectly estimated to a certain
25 degree (Cole and Preacher, 2014). Furthermore, we acknowledge that without a
26 full structural model, we cannot account for the measurement error of individual
27 items. Path analysis was used instead due to our relatively small sample size (n
28 = 173). While many others have used path analysis in a similar manner as we
29
30
have (Bong, 2001; Jones et al., 2012; Lent et al., 2008), full structural models are
31 superior for testing and generating theory. The models reported here are merely
32 an estimation of parameters and do not necessarily accurately describe the
33 motivational phenomena at play. Therefore, we encourage others to cross-
34 validate our models using a different sample to ensure that the pattern of
35 relationships observed here is not a capitalization on chance (Hermida, 2015).
36
37 We encourage others to examine additional longitudinal sets of
38 relationships among grades, self-efficacy, interest, and effort beliefs as well as
39 other affective processes such as student engagement. It is possible that effort
40 beliefs or self-efficacy at the end of the semester could be an outcome of mid-
41 term exam scores based on their reciprocal relationship (Williams & Williams,
42
43
2010). Jones et al. (2012) found that mid-term math grades had a positive
44 relationship with incremental theory of intelligence measured later in the year. In
45 addition, there could be possible mediator relationships that we were unable to
46 explore in this study due to the timing of measurements. Gonzales and Paoloni
47 (2015a) found that situational interest was a mediator between indicators of
48
student engagement (effort, persistence, and attention) and performance in
49
50 physics. Student engagement has been linked to numerous motivation-related
51 processes such as persistence, interest, and effort (Ainley, 2012; Skinner,
52 Kinderman, and Furrer, 2009). By staggering the assessments of students’ self-
53 efficacy, interest, and effort beliefs across time, it becomes possible to look at
54 temporally causal and mediator relationships among them. This work is vital for
55
56
understanding a bigger picture of how student motivation develops and evolves
57 over time in introductory chemistry courses.
58
59
60
Chemistry Education Research and Practice Page 18 of 24
View Article Online
DOI: 10.1039/C6RP00148C
1
2
3
Despite these limitations, we believe the work presented here is important
4
5 for the chemical education community by adding to the burgeoning base of
6 research on affective and motivational processes of chemistry students. We have
1
2
3
Blackwell, L., Trzesniewski, K. H., and Dweck, C. S., (2007). Implicit theories of
4
5 intelligence predict achievement across an adolescent transition: A
6 longitudinal study and an intervention. Child development, 78(1), 246-263.
1
2
3
Ferrell, B. and Barbera, J., (2015). Analysis of students' self-efficacy, interest,
4
5 and effort beliefs in general chemistry. Chemistry Education Research and
6 Practice, 16(2), 318-337.
1
2
3
Huck, S., (2012). Reading Statistics and Research (6th ed.). Boston, MA:
4
5 Pearson.
6 Hull-Blanks, Elva, Kurpius, S. E. R., Befort, C., Sollenberger, S., Nicpon, M. F.,
18 Jones, B. D., Wilkins, J. L. M., Long, M. H., and Wang, F. H., (2012). Testing a
19 motivational model of achievement: How students' mathematical beliefs
20 and interests are related to their achievement. European Journal of
21 Psychology of Education, 27(1), 1-20.
22 Jöreskog, K. G., and Sörbom, D., (2006). LISREL 8.80 for Windows. Computer
23
24 Software. Lincolnwood, IL: Scientific Software International, Inc.
25 Kenny, D. A., (1979). Correlation and Causality. New York: Wiley.
26 Landis, R., Edwards, B. D., and Cortina, J., (2009). Correlated residuals among
27 items in the estimation of measurement models. In R. J. Vandenberg and
28 C. E. Lance (Eds.), Statistical and methodological myths and urban
29
30
legends: Doctrine, verity and fable in the organizational and social
31 sciences. New York: Routledge.
32 Larson, L. M., Stephen, A., Bonitz, V. S., and Wu, T. F., (2014). Predicting
33 Science Achievement in India Role of Gender, Self-Efficacy, Interests, and
34 Effort. Journal of Career Assessment, 22(1), 89-101.
35 Lee, W., Lee, M. J., and Bong, M., (2014). Testing interest and self-efficacy as
36
37 predictors of academic self-regulation and achievement. Contemporary
38 educational psychology, 39(2), 86-99.
39 Lent, R. W., Brown, S. D., Brenner, B., Chopra, S. B., Davis, T., Talleyrand, R.,
40 and Suthakaran, V., (2001). The role of contextual supports and barriers in
41 the choice of math/science educational options: A test of social cognitive
42
43
hypotheses. Journal of counseling psychology, 48(4), 474-483.
44 Lent, R. W., Brown, S. D., and Hackett, G., (1994). Toward a unifying social
45 cognitive theory of career and academic interest, choice, and
46 performance. Journal of Vocational Behavior, 45(1), 79-122.
47 Lent, R. W., Brown, S. D., and Larkin, K. C., (1984). Relation of self-efficacy
48
expectations to academic achievement and persistence. Journal of
49
50 counseling psychology, 31(3), 356.
51 Lent, R. W., Brown, S. D., and Larkin, K. C., (1986). Self-efficacy in the prediction
52 of academic performance and perceived career options. Journal of
53 counseling psychology, 33(3), 265-269.
54 Lent, R. W., Sheu, H. B., Singley, D., Schmidt, J. A., Schmidt, L. C., and Gloster,
55
56
C. S., (2008). Longitudinal relations of self-efficacy to outcome
57
58
59
60
Chemistry Education Research and Practice Page 22 of 24
View Article Online
DOI: 10.1039/C6RP00148C
1
2
3
expectations, interests, and major choice goals in engineering students.
4
5 Journal of Vocational Behavior, 73(2), 328-335.
6 Lewis, S. E., and Lewis, J. E., (2007). Predicting at-risk students in general
1
2
3
Pajares, F., and Miller, D. M., (1994). Role of self-efficacy and self-concept
4
5 beliefs in mathematical problem solving: A path analysis. Journal of
6 educational Psychology, 86(2), 193-203.
18 Raykov, T., and Marcoulides, G. A., (2000). A First Course in Structural Equation
19 Modeling. Mahwah, N.J.: Lawrence Erlbaum Associates, Inc.
20 Robbins, S. B., Lauver, K., Le, H., Davis, D., Langley, R., and Carlstrom, A.,
21 (2004). Do psychosocial and study skill factors predict college outcomes?
22 A meta-analysis. Psychological bulletin, 130(2), 261-288.
23
24 Schermelleh-Engel, K., Moosbrugger, H., and Müller, H., (2003). Evaluating the
25 fit of structural equation models: Tests of significance and descriptive
26 goodness-of-fit measures. Methods of psychological research online, 8(2),
27 23-74.
28 Siegel, R. G., Galassi, J. P., and Ware, W. B., (1985). A comparison of two
29
30
models for predicting mathematics performance: Social learning versus
31 math aptitude–anxiety. Journal of counseling psychology, 32(4), 531-538.
32 Silvia, P. J., (2003). Self-efficacy and interest: Experimental studies of optimal
33 incompetence. Journal of Vocational Behavior, 62(2), 237-249.
34 Skinner, E., Kinderman, T., & Furrer, C. (2009). A motivational perspective on
35 engagement and disaffection. Conceptualization and assessment of
36
37 children’s behavioral and emotional participation in the academic activities
38 in the classroom. Educational and Psychological Measurement, 69(3),
39 493-525.
40 Smist, J. M., (1993). General Chemistry and Self-Efficacy. Paper presented at
41 the Paper presented at the American Chemical Society national meeting,
42
43
Chicago, IL.
44 Smith, P. L., and Fouad, N. A., (1999). Subject-matter specificity of self-efficacy,
45 outcome expectancies, interests, and goals: Implications for the social–
46 cognitive model. Journal of counseling psychology, 46(4), 461-471.
47 Sorich, L., and Dweck, C. S. (1997). Reliability data for new scales measuring
48
students' beliefs about effort and responses to failure. Unpublished raw
49
50 data. Columbia University.
51 Spencer, H. E., (1996). Mathematical SAT test scores and college chemistry
52 grades. Journal of Chemical Education, 73(12), 1150-1153.
53 Steiger, J. H., (1990). Structural model evaluation and modification: An interval
54 estimation approach. Multivariate behavioral research, 25(2), 173-180.
55
56
Streiner, D. L., (2005). Finding our way: an introduction to path analysis.
57 Canadian Journal of Psychiatry, 50(2), 115-122.
58
59
60
Chemistry Education Research and Practice Page 24 of 24
View Article Online
DOI: 10.1039/C6RP00148C
1
2
3
Streiner, D. L., (2006). Building a Better Model: An Introduction to Structural
4
5 Equation Modelling. Canadian Journal of Psychiatry, 51(5), 317-324.
6 Tempelaar, D. T., Rienties, B., Giesbers, B., and Gijselaers, W. H., (2015). The