You are on page 1of 18

PRIOR KNOWLEDGE,ATTITUDE,

AND STRATEGYUSE IN AN
INTRODUCTIONTO
STATISTICSCOURSE

PAUL A. SCHUTZ
UNIVERSITY OF GEORGIA,ATHENS

LISA M. DROGOSZ
UNIVERSITY OF AKRON

VICKI E. WHITE
UNIVERSITY OF AKRON

CHRIS DISTEFANO
UNIVERSITY OF GEORGIA,ATHENS

ABSTRACT: The present study used both quantitative and qualitative methods
to investigate the learning and motivational strategies used by students in a be-
ginning-level statistics course. The research questions that guided the investiga-
tion are: (1) Do motivational variables account for unique variance in the aca-
demic performance of statistics students? (2) Do deeper-level processing
strategies account for unique variance in the academic performance of statistics
students? (3) Do successful students report using different motivation and leam-
ing strategies than unsuccessful students in a beginning-level statistics course?

Direct all correspondence lo: P. A. Schutz, Department of EducationalPsychology, University of Georgia, 325 Aderhold Hall, Athens, GA 30602-
7143. E-mail: pschuh@coe.uga.edu

Learning and Individual Dillerences, Volume 10, Number 4,1998. pages 291-308. Copyright 0 1999 by Elwier Science Inc.
All rights of reproduction in any form reserved. ISSN: 1041-6080
292 LEARNINGANDlNDlVlDUAL DIFFERENCES VOLUMElO,NUMBER4,1998

Ninety-four students enrolled in six sections of the same course over a two-year
period completed measures designed to assess attitudes about statistics, motiva-
tion and learning strategies use as well as previous math and statistics knowl-
edge. In addition, randomly selected participants were interviewed about how
they prepared for their first test. The results of the study show that motivation
variables influenced performance in the introduction to statistics class. In addi-
tion, there was mixed support for the use of deeper-level processing strategies. In
the regression analysis, the learning strategies did not account for unique vari-
ance, but in the interviews and the cluster analysis, their use tended to be related
to performance in the course. Finally, there were differences in the patterns of re-
ported use of different motivation and learning strategies between successful and
unsuccessful students in a beginning-level statistics course. These results help to
expand our understanding of what is involved in the process of learning statistics.
In addition, suggestions for teaching statistics are explored.

The required introduction to statistics course presents interesting challenges for


students as well as instructors. It is a course that normally involves students from
a variety of academic disciplines and levels of statistical and mathematical skills.
For some students, it is the course about which their peers have warned them.
They have heard the horror stories and believed them. For these students, the
course can become the focal point for test, statistics, and computer anxieties. On
the other end of the continuum are the students for which the course is more of a
review. They may have had similar courses or they may have a strong math back-
ground. In any event, they have few problems while taking the class.
The concern of instructors who teach statistics classes,in part, stems from their
belief in the importance of the course. Instructors see the course as critical to stu-
dents’ success in school and increasingly critical for their professional success.
This belief has resulted in the course being a required part of many graduate and
undergraduate academic programs. On the other hand, the instructors also know
that it is hard for learning to occur when emotions and attitudes in the classroom
are ranging from boredom for some students to frustration, anger, and fear for
others.
These and other challenges involved in the teaching of statistics have resulted
in a variety of studies designed to investigate factors that influence successin sta-
tistics classes.Some of the research has shown how students’ prior mathematics
skills positively influence success (Elmore & Vasu 1980, 1986; Elmore, Lewis, &
Bay 1993; Feinberg & Halperin 1978; Presley & Huberty 1988; Woehlke & Leitner
1980). Other research has shown how positive attitudes toward statistics appears
to contribute to successin statistics courses (Elmore et al. 1993; Green 1994; Rob-
erts & Saxe 1982; Wise 1985). In addition, test and statistics anxiety has been
shown to decrease performance in a statistics course (Benson 1989).
Finally, others have looked at how the gender of the student, feminist atti-
tudes, and spatial ability influence performance in statistics classes (Elmore &
Vasu 1986; Woehlke & Leitner 1980). These results have been somewhat mixed in
that Woehlke and Leitner (1980) found no gender difference and Elmore and
KNOWLEOGE,ATTlTUDE,ANDSTt?ATEGYUSE/NSTATlSTlCSCOUFfSE 293

Vasu (1986) found a significant performance advantage for their female partici-
pants in the statistics classes.
While this research has informed us about some of the factors that lead to suc-
cess in statistics courses, questions still remain. For instance, research in educa-
tional psychology has demonstrated the importance of motivation and learning
strategies to academic performance (McKeachie, Pintrich, & Lin 1985; Schutz &
Lanehart 1994; Weinstein & Mayer 1985; Wittrock 1990). This research has shown
that motivational factors like expectancy beliefs or confidence about being able to
learn (Bandura 1986; Eccles 1983; Wigfield & Eccles 1992), and beliefs related to
feeling in control of one’s learning (de Charms 1968, 1984; Weiner 1986) are re-
lated to success in the classroom. In addition, learning strategies that involve gen-
erating connections between one’s prior knowledge and new information, such as
elaboration, have been found to be more effective than simple repetition or re-
hearsal strategies for most learning tasks (McKeachie et al. 1985; Weinstein &
Mayer 1986; Wittrock 1990).
Thus, even though there is research on the factors that influence performance
in statistics courses and the use of motivation and learning strategies in other ar-
eas, these issues have not, to our knowledge, been combined. To investigate these
issues, the present study was designed to look at the motivation and learning
strategies used by students in an introduction to statistics course. The following
research questions were posed to guide the study:
1. Do motivational variables account for unique variance in the academic perfor-
mance of statistics students?
2. Do elaboration learning strategies account for unique variance in the academic
performance of statistics students?
3. Do successful students report using different motivation and learning strate-
gies than unsuccessful students in a beginning-level statistics course?

METHODS

PARTICIPANTS
The participants were 94 graduate students who enrolled in six sections of a
beginning-level statistics course over a two-year period. Seventy-eight percent of
the participants were women. The students represented graduate programs in
nursing, counseling psychology, school psychology, and education. The partici-
pants were Ph.D., Ed.D., and Masters degree candidates.

INSTRUMENTS
A 30-item math test was used to assess the participants’ prior math knowledge.
Mathematics skills that were tested included the use of exponents, multiplication,
division of fractions, and other basic algebra applications. The math problems
294 LEARNING AND INDIVIDUAL DIFFERENCES V0LlJME10,NUMBER4.1998

were similar to the operations the participants would use to compute the mean,
standard deviation, t-test, F-test, and other commonly used equations in the
course (coefficient alpha = .86).
The participants’ prior statistical knowledge was measured using a fill-in tree
diagram (Naveh-Benjamine & Lin 1991). Research in university classes has dem-
onstrated how this method differentiated between students of different levels of
ability (Naveh-Benjamine, McKeachie, Lin, & Tucker 1986; Naveh-Benjamine,
McKeachie, & Lin 1989). In this particular study, a fill-in tree diagram was devel-
oped by the researchers and was designed to measure the students’ prior knowl-
edge of statistical terms and the relationships among these statistical terms
(Naveh-Benjamine & Lin 1991). The fill-in tree was scored by giving one point for
each concept that was placed in the keyed spot on the fill-in tree. In other words, a
high score on this measure reflected a better understanding of the terms to be
used in the class and the relationships among those statistical terms (coefficient
alpha = .77).
The measure for students’ attitudes toward statistics was the Survey of Atti-
tudes Towards Statistics (SATS) (Schau, Dauphinee, & Del Vecchio 1993). The
SATS is made up of four scales: affect, cognitive competence, value, and diffi-
culty. The survey uses a 7-point Likert scale that ranges from 1 (strongly disagree)
to 7 (strongly agree). The scales selected for this analysis were the affect and value
scales. These scales were selected because they represent the students’ attitudes
toward statistics that some research has shown to be important for statistics
achievement (Elmore et al. 1993; Green 1994; Roberts & Saxe 1982; Schau et al.
1993; Wise 1985).
The affect scale of the SATS consists of six items with a reported coefficient al-
pha of .83. For this study, the coefficient alpha was .82. This scale reflects the de-
gree of positive feelings concerning statistics (Schau et al. 1993). Sample items
from the affect scale are: “I will like statistics” and “I am scared by statistics” (re-
verse scored).
The value scale consists of nine items with a reported coefficient alpha of .83
(Schau et al. 1993). For this study, the coefficient alpha was .81. This scale reflects
attitudes about the usefulness, relevance, and worth of statistics in personal and
professional life (Schau et al. 1993). Sample items from the value scale are: “Statis-
tics is worthless” (reverse scored) and “Statistics skills will make me more em-
ployable.”
In addition, several items were added to the end of the SATS to assess the par-
ticipants’ expected grade and confidence in mastering statistics. Of particular in-
terest for this analysis was the question regarding the students’ confidence in
mastering statistics: “How confident are you that you can master introduction to
statistics material?” This question was used as a motivational variable that mea-
sured the participants’ expectancy beliefs about being able to master statistics (Ec-
cles 1983; Wigfield & Eccles 1992). This scale ranged from 1 (not at all confident)
to 7 (very confident).
To measure other motivation and learning strategies used by the participants,
the Motivated Strategies for Learning Questionnaire (MSLQ) was used (Pintrich,
Smith, Garcia, & McKeachie 1991). The MSLQ is an H-item instrument that was
KNOWLEDGE,ATTiTUDE,ANDSTRATEGYUSEINSTAT/ST/CSCOUi?SE 295

developed for and is mainly used with undergraduate students. The MSLQ is
made up of both motivation scales (e.g., intrinsic goal orientation, extrinsic goal
orientation, and control of learning beliefs) and learning strategies scales (e.g.,
elaboration, time and study environment management, and effort regulation). In
addition, the MSLQ has a test anxiety scale that was included. The potential re-
sponses range from 1 (not at all true of me) to 7 (very true of me). The scales se-
lected for this analysis were control of learning beliefs, elaboration, and test anxi-
ety. These scales were selected because they represent motivation (i.e., control of
learning), cognitive learning strategies (i.e., elaboration), and emotions (i.e., test
anxiety) that research has shown to be important for student learning (Pintrich &
De Groot 1990; McKeachie 1990; Schutz & Lanehart 1994; Weinstein & Mayer
1985).
The control of learning beliefs scale consists of four items with a reported coef-
ficient alpha of .68 (Pintrich et al. 1991). For this study, the coefficient alpha was
.70. This scale measures whether or not students feel they are in control of their
learning in a course. Sample items from the control of learning beliefs scale are:
“If I study in appropriate ways, then I will be able to learn the material in this
course” and “If I don’t understand the course material, it is because I didn’t try
hard enough.”
The elaboration scale consists of six items with a reported coefficient alpha of
.76 (Pintrich et al. 1991). For this study, the coefficient alpha was .74. This scale
measures how well students can create elaborations to foster deeper-level pro-
cessing strategies. Sample items from the elaboration scale are: “When I study for
this class, I pull together information from different sources, such as lectures,
readings and discussions” and “I try to relate ideas in this subject to those in other
courses whenever possible.”
The test anxiety scale consists of five items with a reported coefficient alpha of
.80 (Pintrich et al. 1991). For this study, the coefficient alpha was .86. This scale
measures whether or not students experience test anxiety while taking an exam.
For this scale, a high score means low test anxiety. Sample items from the test anx-
iety scale are: “When I take tests I think of the consequences of failing” and “I
have an uneasy, upset feeling when I take an exam.”
The measure for academic performance was the participants’ course grades.
The grades were assigned based on the students’ performance on two tests (each
worth 30% of the grade), computer assignments (worth 10% of the grade), and a
paper (worth 30% of the grade). Course grades were assigned based on a 12-point
scale (e.g., A = 11, A- = 10, B+ = 9, etc.). Because we were not interested in com-
paring across the different sections of the class, all variables used in the analysis
were standardized to a mean of 0 and a standard deviation of 1 within each class.

PROCEDURE
Students were informed at the beginning of the course that we were conduct-
ing a study on how people learn statistics. They signed informed consent forms
and were told that data would be collected at the beginning and the end of the
course. The pre-math and pre-statistics measures used in the study were given
296 LEARNlNGANDlNDlVlDUAL DIFFERENCES VOLUMElO,NUMBER4,1998

during the first day of class, along with a demographic information sheet. The
other measures discussed in this paper were administered during the next-to-the-
last class meeting prior to the second test.
During the final term of the study, interviews were conducted with nine stu-
dents to gather more detailed information to answer the research question: “Do
successful students report using different strategies than unsuccessful students in
a beginning-level statistics course ?” A list was made of the seven highest per-
formers (high performers) and the seven lowest performers (low performers) out
of the 30 participants that term. Four people were randomly selected from each
list, three interviewees and one alternate. All three of the participants on the high-
performer list agreed to participate. Due to a time constraint, one person selected
from the low-performer list was unable to participate, so the alternate was used.
In addition to these six students, three others were selected who had low scores
on the pre-math and pre-statistics measures given at the beginning of the semes-
ter but who did relatively well on the first test (low prior-knowledge high per-
formers). We were interested in this latter group because previous research
showed that success in statistics was related to prior math and statistics knowl-
edge. These students did well on the first test even though they scored lower on
the measures that in the past were shown to be related to success in statistics.
Thus, the strategies they reported using were of interest.
Each participant was interviewed for 20 to 30 minutes. All participants were
asked, among other things, to recall a specific study session they had for the first
test and to report, in as much detail as possible, everything they did during that
study session. Other specific questions included: “Did you do any sample prob-
lems outside of class?” ” What sorts of things did you think about as you were do-
ing the sample problems?” “ How did you use your notes for the test?” “How did
you use your book for the test?” “What was the most effective strategy you
used?” “What was the least effective strategy you used?” Interviewees were
probed for specific responses to the questions.

TABLE 1
Correlation Table for the Variables used in the Analyses (n = 94)

1 2 3 4 5 6 7 8 9 10

1) Pre-statistics 1.00
2) Pre-math .w 1.00
3) Test anxiety .Ol .23* 1.00
4) Affect .Ol .38** .51** 1.00
5) Value .Ol .13 -.04 .30** 1.00
6) Confidence .I2 .16 .20* .51”” .44** 1.00
7) Control .09 .06 .ll .44x* .46*” .45*+ 1.00
8) Elaboration -.17* .03 .02 .12 .37** .12 .15 1.00
9) Course grade .32’* .44** .20* .21* .16 .48”* .27’* -.03 1.00

Means 7.06 22.63 4.41 4.39 5.70 5.83 5.53 5.64 9.62
Std. Dev. 3.09 5.08 1.50 1.33 .93 1.25 .94 .87 1.24

Notes: *p < .05; ** p < .Ol


KNOWLEDGE,ATTITUDE,ANDSTRATEGYUSE/NSTATlSTlCSCOURSE 297

RESULTS
Table 1 displays the correlations between the variables used in the analyses. In
terms of performance in the class, it shows that pre-statistics (r = .32, p < .Ol, r* =
.lO), pre-math (r = .44, p < .Ol, r* = .19), and test anxiety (r = .20, p < .05, P = .04)
were, as previous research has shown, significantly related to performance. For
the attitude toward statistics variables, the affect variable that measures positive
feelings concerning statistics was significant (r = .21, ~7< .05, r* = .04), but the
value statistics attitude scale was not significant (Y = .16, p = ns). Both are vari-
ables that past research has shown to be related to performance in statistics
classes.
In addition, both motivational variables, confidence about learning statistics (r =
.48, p < .Ol, r2 = .23) and control of learning beliefs (r = .27, p < .Ol, r2 = .07), were
significantly related to success. The elaboration learning strategies variable was
not significantly related to academic performance (r = - .03, p = ns). Finally, gen-
der differences were examined and found not to be significant in this study and,
therefore, were not included in the other analyses.

HIERARCHICAL REGRESSION ANALYSIS


In order to further investigate the first two research questions, a hierarchical re-
gression analysis was used. Prior to the analysis, two-way interactions involving
all variables and the outcome measure (i.e., test performance) were performed
and no significant interactions were found.
The hierarchical regression analysis was used to determine the total amount of
variance the set of predictors accounted for in the participants’ grade perfor-
mance and to examine how much of the explained variance could be accounted
for by different sets of predictors. Hierarchical analysis is a technique in which
variables are entered into the equation in sets or blocks. The proportion of vari-
ance accounted for by all of the predictor variables is partitioned by noting the in-
crease in the R2value that is due to a particular set of independent variables (Ped-
hazur 1982). The importance of a block of variables depends on its point of entry
into the equation. As such, the question asked is: “How much does this set of pre-
dictor variables contribute to the total amount of variance explained after control-
ling for the more important sets of predictor variables?” A significant increase in
the amount of variance explained is tested by calculating an F-value associated
with the R2 difference test.
For the present study, variables were formed into four blocks based on a com-
bination of theory and prior research. The first block (Block 1) consists of the
background variables (prior statistics knowledge and prior math skills) and test
anxiety. These are variables that previous research has indicated are related to
performance in statistics classes. The second block (Block 2) was the attitude to-
wards statistics. It included the positive affect for statistics and valuing of statis-
tics variables. lhe third (Block 3) block was the motivation variables that included
control of learning beliefs and confidence about mastering statistics. At the last
298 LEARNlNGANDlNDlVlDlJAL DIFFERENCES VOLUME10,NUMEEA4,1998

TABLE 2
Standardized Beta Weights for Variables Predicting Course Grade

Block Added step 1 steQ2 step3 step4

Background variables
Pre-statistics .25* .25* .16a .15
Pre-math .37* .36' .43* .43*
Test anxiety .I1 .13 .14 .14
Attitudinal variables
Affect for statistics -a4 -.34* -34’
Value of statistics .13 -.lO -.08
Motivation variables
Confidence .49' .49*
Control of learning beliefs .2oa .2oa
Learning strategies variable
Elaboration -.05

Degrees of freedom 3,90 5,88 7, 86 8,85


Change in R2 .26 .02 .20 .Ol
F value 10.62 .93 16.03 .29

Notes: ‘significant at the Bonfemoni corrected level; a = p-values that range from the Bonferroni corrected level
to.05.

stage, the elaboration block (Block 4) was added. This block represented the
deeper-level cognitive processing skills. By testing if the amount of variance ex-
plained is significant when the motivational variables block (Block 3) and the
deeper-level processing block (Block 4) are added while also controlling for the
blocks that were previously entered into the equation, the first two research ques-
tions were investigated.
To remain consistent, all comparisons of standardized beta weights for the var-
ious models were conducted with a familywise alpha value of .05. Table 2 pro-
vides some interesting results. The first two blocks were variables that have been
included in previous research on teaching and learning statistics. For Block 1 (i.e.,
pre-statistics knowledge, pre-math skills, and test anxiety), pre-statistics knowl-
edge and pre-math knowledge were significant, but test anxiety combined with
those two variables was not significant. Block 1 accounted for 26% of the variance
in the grades for the participants in the study. Block 2 (attitudinal variables) did
not significantly account for unique variance in the course grade beyond that of
Block 1, yet the variance accounted for by Block 1 decreased to 22% when Block 2
was added. This indicated there was a small amount of shared variance between
the Block 1 and Block 2 variables as related to the course grades.
Block 3 (motivation variables) was added to investigate the first research ques-
tion. Notice there is a 20% gain in the amount of variability explained. For the two
variables, confidence was significantly related to test performance at the adjusted
level and the control of learning beliefs had an actual p-value of .04. Also, when
the motivation variables were added, the affect for statistics variable also became
significant. Finally, when the Block 3 variables were added, the variance ac-
counted for by Block 1 remained at 22% and the variance accounted for by Block 2
was 6%. This again would indicate a small amount of shared variance between
the blocks’ variables as related to the course grades.
KNOWLEDGE,ATTlTLlDE,ANDSTt?ATEGYUSEINSTATlSTlCSCOURSE 299

Lastly, elaboration was added to address the second research question. From
Table 2, adding elaboration to the model only increases the R2 value slightly. Fur-
ther, the regression weight for this variable is non-significant. In addition, when
the Block 4 variables were added, the variance accounted for by Block 1 remained
at 22%, Block 2 remained at 6%, and Block 3 stayed at .20. Thus, this would indi-
cate that there was only a small amount of shared variance between the blocks
variables as related to the course grades. The full model with all variables in-
cluded accounted for 47% of the variance in course grade.

INTERVIEW ANALYSIS
The interviews were analyzed by looking for patterns in the types of strategies
used by the three groups interviewed (i.e., high performers, low performers, and
low prior-knowledge high performers). When presenting the quotations from the
participants in each of the groups, we have included quotations from each of the
participants. Thus, in each section of the interview analysis where we provide
quotations, all three participants are represented. This indicates some level of con-
sistency at least among the three members of each group.
Overall, it appears that members of the three groups employed similar strate-
gies (e.g., memorization, working through problems, reading, highlighting and
studying the text, and going over notes). The two groups that performed well on
the first test (i.e., high performers and the low prior-knowledge high performers)
tended to go beyond these strategies and use self-monitoring, visualization, relat-
ing the material to the real world with personal examples, and trying to keep up
with the material. On the other hand, the group that performed poorly tended not
to do those things.
The high-performer group had a number of strategies in common as they stud-
ied for the exam. First, the people in this group studied on a regular basis rather
than waiting until the last minute. They did this by either reviewing everyday af-
ter class or by staying “mentally in touch” with the material throughout the se-
mester. This is illustrated in the following responses: “My study habits are the
same from day one as they are right before the test day”; “You don’t completely
separate yourself from the material”; ” What I do after class and in my car, I just
think about what the professor was talking about and I’ll just think about it”; and
“Before class on my way to school I’ll think about it.”
Second, there were indications of self-monitoring. For example, self-question-
ing was used: “Do I understand this?“; “I look at the formula and (say) ‘Can I
read the formula? Do I understand it?“‘; and “Pause for a second if you’re not do-
ing anything and (ask), ‘What do I know about this particular thing? What do I re-
member?“’
Third, these participants relied on visualizations of the material. For example: “I
sort of take a mental snap shot of the formula “; “I picture things; I’m really visual. I
picture diagrams and stuff like that (when trying to learn the material)“; “I can pic-
ture my notes”; and “I picture the normal curve (to remember dispersion).”
Fourth, the high performers tended to connect material with real world or per-
sonal examples to remember it. Examples of this included: “Out in the field, how
300 LEARNINGANDINDIVIDUAL DIFFERENCES VOLUMElO,NUMBER4,199E

would this relate to reality?“; “I try to relate (the problems to my research) to see
if I really understand”; and “I try to relate it to other things in other classes . . . to
really integrate (the material) because that way I remember.” In addition to con-
necting the material to personal experiences, they also made attempts to integrate
the class material together. For example: “Sometimes I’m saying to myself, ‘How do
things fit together? How (does) the standard deviation . . . fit in with the t-test”‘;
“If you can’t tie everything into what you already know it can be cumbersome”;
and “I guess you could say I’m trying to relate it back to (other chapters).”
The final type of strategy that differentiated the high performers from the other
two groups was a consistent attempt to redefine the material in their own words.
Examples that demonstrate this are: “I’d rather have the exact wording (when
taking notes from the text) and then I paraphrase in my mind” and “I don’t like to
remember real wordy definitions. I like examples because I always remember the
examples . . . and then (I) come up with my own definitions from the examples.”
The low prior-knowledge high performers also had interesting ways in which
they learned the material. Though most of them had statistics courses in the past,
they tended not to remember much (demonstrated by lower scores on the pre-
math and pre-statistics tests). This group used the same sort of general strategies
as the other groups-reading, highlighting, and memorizing-but engaged in ex-
tra activities to make up for their lack of knowledge. In other words, they may
have been aware they did not know the material so they made a special effort to
look at other sources to help them learn the material. The extra time spent with
the material each week included receiving weekly tutoring, spending a few hours
the day after class to go over the material, and using other textbooks to help clar-
ify the material. These are illustrated by the following responses: “I basically went
through a programmed instruction book and then went back and reread the sta-
tistics book we have”; “(Th e next morning) I’ll put in a good couple of hours to try
to solidify (the concepts), otherwise the longer I go without them, I have to start
all over again”; and “When I would do a chapter from the book we were using, I
would go to (another text) and try to find a chapter that was covering the same
material (and do the problems).”
This group also used some of the same strategies as the high performers. These
included visualization (“I . . . picture it in my mind . . . how the scores are dis-
persed around the mean . . . I have a picture in my mind like actually seeing it”);
self-monitoring (“I’m not only starting to know the steps [to standard deviation]
but to try to understand what it means”); and integration of the material
jhtooking at the picture itself and relating where the mean, median, and mode
are on a normal curve, and then looking at the abnormal [skewed] ones”).
Overall, the low prior-knowledge high performers felt their test scores were
due to hard work and keeping up with the material. When they were questioned
directly about what they attributed their scores to, the following responses were
given: “Keeping up and studying, the repetition, but mostly keeping up”; “I
guess because I really worked hard”; and “A lot of studying.”
The final group of individuals is the low performers. These participants basi-
cally used memorization, read the textbook, and did sample problems, as did the
other two groups. What differentiates this group from those that performed better
KNOWLEOGE,ATTlTUDE,ANDSTRATEGYUSEINSTATlSTlCSCOURSE 301

was the extent to which they used these and other strategies. It appears they did
not move beyond strictly trying to memorize the material to integrating it and ap-
plying it to personal experiences. The difficulties in this group are illustrated by
the following responses to how they prepared for the exam: “I just went over my
notes”; “What I think I got was a literal understanding of the terms, I didn’t have
an in-depth understanding”; ” I wasn’t able to come up with any good . . . integra-
tive types of examples to study”; and “I . . . just looked at the (diagrams in the
book) and reviewed them.”
This group also felt overwhelmed by the material and the time in which they
had to prepare for the exam. Examples that express this concern are: “I felt over-
whelmed, I felt lost”; “I wasn’t prepared”; “I was fatigued”; “I wasn’t sure what I
was supposed to be studying”; and “I didn’t spend the amount of time I should
have studying for that test.”
Interestingly, when comparing the prior math and statistics knowledge of the
groups, the low-performer group actually had higher means overall than the low
prior-knowledge high-performer group on the 30-point pre-math test (M = 20.67
vs. M = 18.67) and the 15-point pre-statistics measure (M = 6.67 vs. M = 5). This
suggests they may have had more, or at least the same, prior knowledge. The
poor performance of the low-performer group may be attributed to a lack of prep-
aration and failure to connect the material. This group, when compared to the low
prior-knowledge high-performer group, may have failed to recognize their need
to work more with the material because they had low prior knowledge. This may
be due to anxiety and/or fear of the material.

PRELIMINARY DISCUSSION
The purpose of this study was to investigate how people learn statistics. This was
done by combining variables that were used in research on success in statistics
classes (Elmore et al. 1993; Elmore & Vasu 1980,1986; Feinberg & Halperin 1978;
Green 1994; Presley & Huberty 1988; Roberts & Saxe 1982; Schau et al. 1993; Wise
1985; Woehlke & Leitner 1980) with variables from research on how motivation
and learning strategies use influence successful learning (McKeachie et al. 1985;
Schutz & Lanehart 1994; Weinstein & Mayer 1986; Wittrock 1990).
The results of the regression analyses indicated the findings from this study
were similar to previous studies that found relationships between pre-statistics
knowledge and pre-math skills and student success in statistics courses. There-
fore, this group of participants tended to perform similarly to those in other stud-
ies where these issues have been investigated. In addition, as anticipated, the two
motivational variables included in the analyses made significant contributions to
the amount of variance accounted for in course performance. This indicates that
students in the course who were confident in their abilities to master the course
material and saw themselves as being in control of what went on in the course
were more successful in the course.
An unexpected result in this analysis was the lack of a significant finding in the
302 LEARNINGANDINDIV~DUAL DIFFERENCES VOLUMElO.NUMBER4.1998

analyses for the elaboration learning strategies scale. This was unexpected in that
prior research indicated a relationship between deeper-level processing strategies
and performance. There was also evidence from the interviews that some of the
participants who did well on the first test tended to develop analogies, relate the
information to their prior knowledge, and create visualization for the to-be-
learned information. In addition, the interview participants who did not do well
on the first test tended to use more rehearsal or repetition strategies. Thus, there
were conflicting findings related to the influence of elaboration learning strategies
on performance in the introduction to statistics classes.
There are a variety of potential explanations as to why these unexpected find-
ings occurred. For example, there may be no relationship to find. Thus, although
the relationship between elaboration and learning has been established in other
areas, it may not hold in an introduction to statistics class. There could also be
measurement problems. For example, there may be a relationship, but because of
reliability and validity issues with the measure used, we did not find the relation-
ship. In other words, the concept of deeper-level processing may be broader than
what is actually measured by the elaboration learning strategies scale.
An additional explanation for these unexpected findings is the possibility that
there were groups or clusters of participants who, for example, did not report us-
ing elaboration strategies at a high level yet did well in the class, or who did not
report valuing statistics yet did well in the class. If such groups existed, they may
have the effect of lowering the correlation between elaboration and performance
or between valuing statistics and performance. Thus, in order to investigate this
possibility and to try to identify clusters of students who showed different pat-
terns of prior knowledge, attitudes towards statistics, and motivation and learn-
ing strategies use during an introduction to statistics class, we performed a cluster
analysis.

ADDITIONAL ANALYSIS
By using the cluster analysis, the attempt was to tease apart the relative influ-
ences of different variables on academic performance and identify clusters of stu-
dents who showed different patterns of prior knowledge, attitudes towards statis-
tics, and motivation and learning strategies use during an introduction to
statistics class. Cluster analysis is associated with a wide variety of classification
techniques that begin with a single data set and attempt to organize the cases in
the data set into homogeneous groups based on the information provided by the
variables (Aldenderfer & Blashfield 1985).
To separate the data into groups, a two-step procedure was employed that
combined the Wards method and the K-means algorithm method. Wards hierar-
chical method begins with each case as a separate cluster and successively groups
similar clusters together to create clusters of multiple cases. The Wards method is
designed to minimize the variance within the clusters. Beginning from the posi-
tion where each case is its own cluster, cases are joined by finding those that con-
KNOWLEDGE, ATTITUDE, AND STRATEGY USE IN STATlSTlCS COURSE 303

tribute the smallest amount to the error sum of squares. Once a case is assigned to
a specific cluster, it becomes a permanent member of that cluster.
The second method, K-means algorithm method, belongs to the family of itera-
tive partitioning methods. By making more than one pass through the data, this
algorithm may be able to overcome a poor initial partition. For this analysis, a
Wards method analysis was conducted and the final centroid results noted. These
resulting centroids from the Wards method analysis serve as the initial partitions,
or “seeds”, for the K-means analysis. By using a combination of the two methods,
the researchers are entitled to the benefits of both.
Initially, a plot of the cubic clustering criterion by number of clusters was used
to determine the number of clusters underlying the data set. From this plot, two-
to six-cluster solutions were run and analyzed. A four-cluster result was settled
on based on the researchers’ judgment and the interpretability of the solution. In
addition, we consider these cluster groups to be tentative because of the small
number of participants relative to the number of variables used in the analysis.
For this analysis, the variables used were the prior knowledge variables (i.e., pre-
statistics and pre-math), the attitude variables (i.e., positive affect and value sta-
tistics), the motivation variables (i.e., confidence and control of learning), the
learning strategies variable (i.e., elaboration), and the emotion variable (i.e., test
anxiety).

CLUSTER ANALYSIS RESULTS


Table 3 illustrates the mean differences between the groups on the variables
used in the cluster analysis and test perf0rmance.L Cluster 1 (20 participants) rep-
resents some of the successful statistics students. These students have high scores
on most of the variables. They have a strong background in math, prior statistics
knowledge, enjoy the subject, feel they could master statistics, and have low test-

TABLE 3
Mean Differences Between the Cluster Groups on Variables Used in the Cluster Analysis and
Course Grade*

Cluster 1 Cluster 2 Cluster 3 Cluster 4


(n = 20) (n = 24) (n = 24) (n = 26)

Pre-statistics l.lP - .59b** -.19b -.15b*”


Pre-math .44a .41@ .58b -.1&+-b

Test anxiety .2tW .6Sa -.85c -.05b


Affect .51a .9oa -.62b -.65b
Value - .05b .52” ,598 - .98c
Confidence ,498 .54++ .13a** -l.OOb
Control of learning .44a .51a .14a - .94b
Elaboration -.7P .49= .5oa - .32b
- ,‘J,b”*c
Course grade .6Oa .3lab*’ - .54c

Notes: *Means with different superscripts are significantly different using the Bonferroni method with a family-
wise alpha level of .05. Means with the same superscripts and having an ** have p-values that range from
.0125 to .05.
304 LEARNlNGANDINDII/IDUAL DIFFERENCES VOLUMElO,NUMBER4,1998

anxiety scores. On the other hand, these students tended to report not using elab-
oration as a learning strategy when studying. This cluster of students had the
highest test scores. Therefore, this seems to be a group of students for which the
class may have been, at least in part, a review.
Cluster 2 had 24 students. This grouping of students was well above the mean
on pre-math and test anxiety, yet they were well below the mean on the pre-statis-
tics measure. This group of students was also above the average on the attitude,
motivation, and elaboration variables. Finally, these students had the second
highest average grade in the course. It was concluded that this was a group of stu-
dents who did well in the class and reported using elaboration learning strategies.
If we compare Clusters 1 and 2, we see that Cluster 1 students were signifi-
cantly higher than Cluster 2 students on prior statistics knowledge, but that Clus-
ter 2 was significantly higher than Cluster 1 on valuing statistics and the elabora-
tion variables. Cluster 2 was similar to Cluster 1 on math skills, test anxiety, affect,
confidence, and control of learning beliefs. In terms of course grade, Cluster 1 had
a higher (but not significantly different) average grade than Cluster 2. Thus, it
would seem that Cluster 2 students may have been able to use elaboration strate-
gies and their valuing of statistics to compensate for their lack of prior knowledge
in statistics. It seems that in comparison to Cluster 1, for which much of the class
was a review, Cluster 2 seems to be a group that gained a significant amount of
knowledge in the class.
The third cluster consisted of 24 students. This group of students thought sta-
tistics was valuable, but they did not like the subject. They had low prior math
knowledge and low prior statistics knowledge. These students reported using
elaboration more than the students in Cluster 2, which would also have the effect
of lowering the correlation between elaboration and performance for all partici-
pants (see Table 1); however, without a math or statistics background with which
to relate the new information to, it did not seem to provide much help. They may
have been trying to use elaboration as a learning strategy before they fully under-
stood the concepts they were trying to elaborate. Elaboration for this group may
have been further problematic because they had the highest test-anxiety score of
the four clusters. Although they saw the value of the course, their lack of prior
knowledge combined with their high test anxiety may have rendered the use of
the elaboration strategies ineffective.
Comparing Cluster 2 to Cluster 3, Cluster 3 had a significantly lower math
background and lower positive affect towards the course. Additionally, students
in Cluster 3 had higher test anxiety and received lower grades than students in
Cluster 2. Thus, although Clusters 2 and 3 were similar on pre-statistics knowl-
edge, value, confidence, control of learning, and elaboration, they were signifi-
cantly different on some of the key variables associated with success in the regres-
sion analysis (e.g., pre-math knowledge and test anxiety).
The final cluster of 26 people could be termed an unsuccessful statistics group.
These students were low on the majority of the variables and had the lowest
grades in the course. These students had, by far, the lowest value, confidence, and
control of learning.
In comparing Cluster 3 to Cluster 4 students, Cluster 4 students came into the
KNOWLEDGE,ATTlTUDE,ANDSTRATEGYUSEINSTATlST/CSCOURSE 305

class with similar prior knowledge of statistics and math as did Cluster 3 stu-
dents, but the students in Cluster 4 were lower on other ey variables such as
value, confidence, control of learning beliefs, and elaboration. So, even though
neither Cluster 3 nor Cluster 4 did very well in the class, Cluster 4 was signifi-
cantly lower on the attitude, motivation, and learning strategies variables and
they also averaged lower grades than Cluster 3 and the other clusters.

DISCUSSION
A working knowledge of statistics is becoming increasingly important in most
professions. Unfortunately, as most instructors know, some students have devel-
oped a fear of statistics. Even without the fear that is sometimes involved in the
course, the subject matter can be difficult. Therefore, it is important that instruc-
tors of statistics be able to facilitate students during the process of learning statis-
tics. In order for instructors to do that, we need a better understanding of the fac-
tors that influence the process of learning statistics.
In this study, we were able to account for 47% of the variance in course perfor-
mance. As previous research has shown, prior statistics and math knowledge
tended to be important in helping to understand the concepts in this introduction
to statistics class. This study also provides some support for the inclusion of moti-
vation and learning strategies variables into the discussion of how people learn
statistics. By including these variables, we were able to get a clearer, but admit-
tedly more complex, understanding of the transactions among the factors in-
volved in learning statistics.
The use of interviews and the cluster analysis also provided a beneficial perspec-
tive on the process of learning statistics. From the interviews, we found that suc-
cessful students on the first test tended to use deeper-level processing strategies.
This type of processing is important because if one plans to use the material in the
future, it has to become personally meaningful. Therefore, this may be an area
where the instructor can facilitate the process by asking questions in class that re-
quire the students to make these connections. It was not that the successful partici-
pants being interviewed did not use rehearsal as a strategy, but they tended not to
stop there. On the other hand, there was little evidence of deeper-level processing
strategies used by the participants who did not do well on the first test. Their strate-
gies tended to be confined to rehearsal- or repetition-type strategies.
A second issue that came out of the interviews was the importance of keeping
up with the material. From the interviews, it was clear the students who did well
also tended to be the ones who indicated they kept up with the material. Since the
class met only once a week, it was important for the students to stay in touch with
the material on their own. The students who did well tended to do that. For the
students who did not do as well, there was little or no evidence of their attempts
to keep up with the material. Also related to this factor was the self-monitoring
and self-regulating that was exemplified by those who were successful in the
class. The participants who did well also tended to provide evidence of monitor-
306 LEARNlNGANDINDI!JlDUAL DIFFERENCES VOLUME10.NUMBEA4.1998

ing their understanding and recall of the material that was not present in the in-
terviews of the less successful participants.
What was clear from the participants who did not do as well on the first test as
the successful participants was a feeling of being lost or overwhelmed by the ma-
terial and the class. This occurred even though the participants had equal or
higher means on the pre-math and pre-statistics tests as the group that had low
prior-knowledge but did well on the test. Thus, one group may have dealt with
their lack of prior knowledge by putting in more effort and using more effective
learning strategies than the other group that did not. On the other hand, the re-
sults of the interview should be considered exploratory. They were conducted
with nine participants for one test. Still, these exploratory results with interviews
are encouraging and indicate a need for additional in-depth interviews.
Although the cluster analysis involved a limited number of subjects per vari-
able (approximately 12 subjects per variable), there seems to be some preliminary
evidence for clusters of students with distinctive patterns of how they approach
the course. This information is important for those who teach statistics because if
these students can be identified early in the term, help can be more specifically di-
rected to the problem area. For example, one key area is the importance of math
knowledge. In most situations, a brief math test could be developed to identify
students who are having problems so that additional math training or math re-
view can be provided and thus may help them gain or regain confidence in their
math skills. Thus, when formulas are used to help students understand a concept,
they are not so focused on the formula that they do not learn the concept.
For example, the students in Clusters 3 and 4 (see Table 3) could be identified
early in the term and be provided review and preview experience with math and
statistics. These experiences can help them be more confident in their math and
statistics background, feel more in control of their efforts, and provide opportuni-
ties for early successes in the course. This, in turn, may help them to deal with
some of their anxieties towards the course. In other words, we know, based on
our experiences teaching introduction to statistics classes, there are students who
need extra help related to their math and statistics background. Therefore, the
earlier they are identified and provided with needed assistance, the better off the
students and the instructor will be.
Ideally it may be nice to teach a statistics course that involved only students
from Clusters 1 and 2. On the other hand, in the real world, all clusters will be
represented in any particular class. Therefore, in addition to identifying students
based on their pre-math, pre-statistics, and even pre-test anxiety towards the
class, it may also be useful to use the information to create heterogeneous groups
for in-class activities. It seems that students from Cluster 4, who were below aver-
age on all the variables, may benefit from interacting with students from Cluster 2
who were interested in and valued the course. In addition, students from Cluster
1 who have good math skills but see little value in statistics may benefit from in-
teracting with students in Cluster 3 who have a high value for statistics but have a
weak math background.
In any event, this research does provide support for the inclusion of motivation
and learning strategies variables into the investigation of how people learn statis-
KNOWLEDGE,ATTlTUDE,ANDSTRATEGYUSEINSTATlSTlCSCOURSE 307

tics. So, even though we consider the results of the cluster analysis to be prelimi-
nary, they do indicate an important avenue for research on how students learn to
use statistics and how instructors can better facilitate that process.

REFERENCES
Aldenderfer, M. S. & R. K. Blashfield. (1985). Cluster analysis. Beverly Hills, CA: Sage.
Bandura, A. (1986). Social foundations of thought and action: A socialcognitive theory. Engle-
wood Cliffs, NJ: Prentice-Hall.
Benson, J. (1989). “Structural components of statistical test anxiety in adults: An explor-
atory model.” The Journal of Experimental Education, 57,247-261.
de Charms, R. (1968). Personal causation: The internal affective determinants of behavior. New
York: Academic Press.
de Charms, R. (1984). “Motivation enhancement in educational setting.” Pp. 275-310 in Re-
search on motivation in education (vol. l), edited by R Ames & C. Ames. New York: Aca-
demic Press.
Eccles, J. (1983). “Expectancies, values and academic behaviors.” Pp. 75-146 in Achievement
and achievement motives, edited by J. T. Spence. San Francisco: Freeman.
Elmore, I’. B. & E. S. Vasu. (1980). “Relationship between selected variables and statistical
achievement: Building a theoretical model.” Journal of Educational Psychology, 72,457467.
Elmore, I’. B. & E. S. Vasu. (1986). “A model of statistics achievement using spatial ability,
feminist attitudes and mathematics-related variables as predictors.” Educational and
Psychological Measurement, 46,215-222.
Elmore, P. B., E. L. Lewis, & M. L. Bay. (1993). Statistics achievement: A function of atti-
tudes and related experience. Paper presented at the annual meeting of the American
Educational Research Association, Atlanta, GA.
Feinberg, L. B. & S. Halperin. (1978). “Affective and cognitive correlates of course perfor-
mance on introductory statistics.” Journal of Experimental Education, 46,11-18.
Green, K. E. (1994). Affective components of attitude and statistics instruction. Paper pre-
sented at the annual meeting of the American Educational Research Association, New
Orleans, LA.
McKeachie, W. J. (1990). “Learning, thinking and Thorndike.” Educational Psychologist, 25,
127-144.
McKeachie, W. J., P. R. Pintrich, & Y.-G. Lin. (1985). “Teaching learning strategies.” Educa-
tional Psychologist, 20,153-161.
Naveh-Benjamine, M. &Y.-G. Lin. (1991). Assessing students’ organization of concepts: A man-
ual for measuring course-specific knowledge structures. Ann Arbor, MI: National Center
for Research to Improve Postsecondary Teaching and Learning.
Naveh-Benjamine, M., W. J. McKeachie, Y.-G. Lin, & T. Tucker. (1986). “Inferring stu-
dents’cognitive structures and their development using the ‘ordered tree technique. ”
Journal of Educational Psychology, 78,130-140.
Naveh-Benjamine, M., W. J. McKeachie, & Y.-G. Lin. (1989). “Use of the ordered-tree tech-
nique to assess students’ initial knowledge and conceptual learning.” Teaching of Psy-
chology, 16,182-187.
Pedhazur, E. J. (1982). Multiple regression in behavioral research: Explanation and prediction.
(2nd Ed). New York: Holt, Rinehart, &Winston.
308 LEARNlNGANDINDIVlDUAL DIFFERENCES VOLUMElO.NUMBER4.1998

Pintrich, P. R. & E. V. De Groot. (1990). “Motivation and self-regulated learning compo-


nents of classroom academic performance.“ Journal of Educational Psychology, 82,334O.
Pintrich, P. R., D. A. F. Smith, T. Garcia, & W. J. McKeachie. (1991). A manual for the use of
fhe Motivated Strategies for Learning Questionnaire. Ann Arbor, MI: National Center for
Research to Improve Postsecondary Teaching and Learning.
Presley, R. J. & C. Huberty. (1988). “Predicting statistics achievement: A prototypical re-
gression analysis.” Multiple Linear Regression Viewpoints, 16,36-77.
Roberts, D. M. & J. E. Saxe. (1982). “Validity of a statistics attitude survey: A follow-up
study.” Educational and Psychological Measurement, 42,907-912.
Schau, C., T. Dauphinee, & A. Del Vecchio. (1993). Evaluation of two surveys measuring
students’ attitudes towards statistics. Paper presented at the annual meeting of the
American Educational Research Association, Atlanta, GA.
Schutz, I’. A. & S. L. Lanehart. (1994). “Long-term educational goals, subgoals, learning
strategies use and the academic performance of college students.” Learning and Individ-
ual Difierences, 6,399412.
Weiner, B. (1986). An attributional theory of achievement motivation and emotion. New York:
Springer-Verlag.
Weinstein, C. E. & R. E. Mayer. (1986). “The teaching of learning strategies.” Pp. 315-327 in
Handbook of research on teaching, edited by M. Wittrock. New York: Macmillan.
Wigfield A. & J. Eccles. (1992). “The development of achievement task values: A theoretical
analysis.” Developmen fal Review, 12,26.5-310.
Wise, S. L. (1985). “The development and validation of a scale measuring attitudes toward
statistics.“ Educational and Psychological Measurement, 45,101-104.
Wittrock, M. C. (1990). “Generative processes of comprehension.” Educational Psychologist,
24,345-376.
Woehlke, P. L. & D. W. Leitner. (1980). “Gender differences in performance on variables re-
lated to achievement in graduate-level educational statistics.” Psychological Reports, 47,
1119-1125.

You might also like