Factors Affecting Student Academic Success in Gateway Courses at Northern Arizona University

Russell Benford Julie Gess-Newsome Center for Science Teaching and Learning Northern Arizona University Flagstaff, AZ 86011-5697

May 24, 2006

Factors Affecting Student Academic Success in Gateway Courses at Northern Arizona University

Table of Contents Section Abstract Introduction Predictors of Student Achievement in Introductory Business, Mathematics, and Science Courses Predictors of Student Achievement in Business, Marketing, and Economics Predictors of Student Achievement in Mathematics Predictors of Student Achievement in Computer Science Predictors of Student Achievement in Physics Predictors of Student Achievement in Chemistry Predictors of Student Achievement in Biology Summary of Factors That Predict Student Success in Introductory Business, Mathematics, and Science Courses Interpreting Results of Predictive Studies in Business, Mathematics, and Science Education Methods Institutional Records and Public Data ABC and DFW Rates in Gateway Courses Characterizing ABC and DFW Students Student Survey Page 4 5 10

11 13 15 18 20 21 24 27 30 31 32 34 35

1

Characterizing Students’ Educational and Socioeconomic Contexts Characterizing Gateway Classrooms and Courses Development of Predictive Model Results Course-Oriented ABC and DFW Statistics ABC and DFW Rates in Gateway Courses Teaching Styles Used in Gateway Courses Student-Oriented ABC and DFW Statistics Student Demographics Student Perception of Course Student Academic Habits Effect of Class on Student Student Perception of College Life and NAU Predictors of Student Success Discussion Summary and Interpretation of Results Conclusions and Recommendations Student Recruitment Student Preparation Student & Faculty Diversity Curriculum Design & Implementation Identification & Intervention Acknowledgements

38 39 40 43 44 44 45 47 49 59 68 78 82 93 99 99 111 112 114 115 118 120 123

2

References Appendix A: High Schools of Origin Appendix B: Survey on Factors Contributing to Student Success Appendix C: Reformed Teaching Observation Protocol

125 140 143 149

3

Factors Affecting Student Academic Success in Gateway Courses at Northern Arizona University

Abstract Students in gateway business, math, and science courses at Northern Arizona University receive non-passing grades (grades of D, F, and W) at high rates. To identify possible trends in demographic groups that receive DFWs and to investigate why students receive DFWs in these courses, a student survey was administered to 719 students in 7 gateway courses, and institutional data were collected on 23255 students enrolled in 15 gateway courses. Student achievement and socioeconomic data on high schools from which gateway students originated were also collected. Student and high school data were analyzed to elucidate differences between ABC and DFW students, and to determine if differences in DFW rates existed between genders and among ethnicities. To determine if instructional style of gateway courses affected DFW rates or patterns in the demographics of DFW distribution, an instrument was used to characterize instructional styles used in the 15 gateway courses. Resulting data were analyzed for trends in DFW rates, gender, and ethnicity. Data suggest that possible causes of DFWs are inadequate student recruitment standards, student academic underpreparedness, lack of student and faculty ethnic and cultural diversity and interaction, and ineffective and inequitable instructional techniques. Possible interventions are discussed.

4

Factors Affecting Student Academic Success in Gateway Courses at Northern Arizona University

Introduction The level of success students achieve in their first semesters of college has far-reaching implications for students’ personal and professional lives. Student success has an immediate influence on a student’s academic selfesteem, persistence in elected majors, and perseverance in higher education. Success in early semesters at college also ultimately impacts students’ postcollege experiences, such as career choice, personal income and level of success, and degree and nature of participation in community life. Thus, the experience a student has in the introductory college classes she or he attends can have a significant influence on the course of that student’s adult life. It is therefore alarming that introductory college classes are among the least enjoyed and least understood classes in a student’s postsecondary academic career. Disaffection with and low performance in introductory college classes is a serious problem at colleges and universities nationwide (Horn et al. 2002, Horn and Premo 1995). The problem is especially evident in introductory business, mathematics, and science courses. Such courses are often required and integral components of an undergraduate education, yet many students who enroll in these courses achieve moderate or low levels of success in them. Low levels of success in introductory business, mathematics, and science courses

5

result in significant attrition of talented students in these areas of study (Gainen 1995, Congress of the United States, Office of Technology Assessment 1988). Attrition in business, mathematics, and science courses does not occur in all demographic groups at an equal rate. Of the major ethnic groups in the United States, African Americans, Hispanics, and Native Americans are less likely to enroll in and more likely to resign from business, mathematics, and science-related majors. Additionally, females are less likely to enroll in and more likely to resign from these courses than are males (Brower and Ketterhagen 2004, National Center for Educational Statistics 2002, Herndon and Moore 2002, Brush 1991, Hilton and Lee 1988). The greatest period of attrition for female students in science-related educational tracks is between the end of high school and the beginning of college (Oakes 1990). When the current employment demographics of science and science-related occupations in the United States are considered (Figures 1 and 2), the notion of undergraduate attrition in the groups that are least well-represented in these areas of employment is disturbing.

6

100 90 80 70 60 50 40 30 20 10 0 Entire Workforce Science & Engineering Occupations Male Female

Figure 1: Gender trends in employment (bachelor’s or higher degree recipients) in the United States (National Science Foundation 2004)

100 90 80 70 60 50 40 30 20 10 0 White Asian Black Hispanic Native American Other

Figure 2: Ethnic trends in science and engineering occupations (bachelor’s or higher degree recipients) in the United States (National Science Foundation 2004)

As these data indicate, student disaffection with and attrition in introductory business, mathematics, and science courses is a national problem. The problem is also, unfortunately, a local one. Levels of student dissatisfaction with and rates of attrition in introductory business, mathematics, and science
7

courses at Northern Arizona University are consistent with national trends (Office of Planning and Research 2003, Horn et al. 2002). Because student satisfaction and perseverance are vital to student success in college, understanding factors that diminish student satisfaction and perseverance is necessary if these problems are to be addressed and overcome. Understanding these factors and implementing administrative changes to address them is especially important in entry-level courses, where student attitudes and habits are fundamentally shaped. Large enrollment, entry-level college courses that are prerequisites for majors or graduation are commonly called “gateway” courses. Students enrolled in gateway courses in business, math, and science at Northern Arizona University (NAU) receive grades of D, F, or W at an alarmingly high rate (mean = 27.1%, SD ± 8.3%*). Such a high DFW rate in gateway courses is of particular concern, because these courses are populated primarily with freshmen and sophomores, and the experiences of these lower division students are likely to affect these students’ personal choices at and after college. It is therefore important to characterize the individuals and groups who have recently received final grades of D, F, or W in these courses, and, if trends in these demographics are apparent, to understand why such individuals and groups have received these grades. Once this is done, a method for identifying individuals who are at increased risk of receiving these grades in the future could be developed, and strategies to help students succeed in these courses could be employed.

8

The percentage of students who receive a final grade of D, F, or W in a course – the DFW rate – is a metric that can be used to gauge a course’s academic success. Assuming grades in the course are awarded for individual merit (opposed to relative standing in the class), a low DFW rate suggests that many students are achieving an acceptable level of competency with the subject matter of the course. Thus, the course is a successful educational endeavor. The interpretation of a course’s DFW rate becomes more complicated, however, when the many factors that can affect the DFW rate are considered. Student factors such as aptitude, motivation, and study habits obviously affect student success. But non-student factors such as the academic environment, course curricula, and pedagogical techniques used by the course instructor can also dramatically affect student success. It is therefore appropriate to also

consider student, teacher, curricular, and environmental influences in concert when interpreting DFW data to evaluate the academic success of a course. Understanding challenges that students face in gateway business, math, and science courses at Northern Arizona University is requisite to helping students achieve a higher level of success in these courses. Greater success is important, because most students enroll in gateway courses at the beginning of their academic careers, and conceptions they form during this period about college life and their own academic skills are lasting. Such conceptions are likely to affect personal, academic, and career choices that students make. Negative conceptions could steer students who perform poorly in gateway courses away from their careers of choice. This change in direction could perpetuate the

9

under-representation of certain groups in business, math, and science professions experienced in the United States today. Thus, the objectives of this study are: 1) to determine who receives DFWs in gateway business, math, and science courses at NAU, 2) to determine why these students receive DFWs in these courses, to 3) to develop a model for identifying students who might be at risk of receiving a D, F, or W in these courses, and 4) to identify and recommend intervention strategies that could improve the rates of academic success in these courses.

* Based on data from ACC256, BA201, BIO100, BIO181, BIO182, CHM151, CHM152, CIS120, ENV101, GLG100, MAT110, MAT125, MAT137, MAT155, PHY111, Fall 2000 through Spring 2002 semesters.

Predictors of Student Achievement in Introductory Business, Mathematics, and Science Courses An abundance of research has been performed in the most recent four decades attempting to identify predictors of student performance in introductory business, mathematics, and science courses. Both cognitive and noncognitive factors have been considered, because numerous studies have shown both types of variables to be useful predictors. Some studies have shown that

noncognitive variables are more useful than cognitive variables in predicting the academic success of nontraditional students (e.g. Sedlacek 2002). In addition to considering numerous types of variables, various methods of data collection and analysis have been used. Varied methods seem appropriate in research on

predictors in business, math, and science because quantitative measures have the potential to overlook the presence and/or magnitude of non-cognitive and

10

qualitative variables (Glesne 1999), and qualitative measures such as freeresponse questionnaires and interviewing are likely to contain biases. For

example, in a meta-analysis of research on variables that contribute to classroom success, McAllister (1996) reports that both teachers and students make “selfserving attributions taking credit for success, but not for failure.” Such biases could result in poorly informed analyses. While some discrepancies among

conclusions from disparate studies exist, overall trends are apparent within each discipline. Furthermore, trends that transcend disciplines are evident, and will be discussed at the end of this review.

Predictors of Student Achievement in Business, Marketing, and Economics

Cognitive and academic variables have been shown to be only adequate predictors of success in introductory business, marketing, and economics courses. Sachdeva and Sterk (1982), Eskew and Faley (1988), Liesz and Reyes (1989), and Doran, Boullion, and Smith (1991) report that locally written and administered placement exams that measure student content knowledge and reasoning skills predict student performance in introductory finance courses. Eckel and Johnson (1983) report that the ACT score in math predicts success in introductory accounting courses. However, some studies contradict this

conclusion and suggest that standardized entrance exam scores are not effective predictors in introductory accounting courses (Brown 1966, Ingram and Peterson 1987).

11

High school and college performance seems to be a more reliable predictor of student success than are entrance exam scores in introductory courses in the business field. Brown (1966) reports that high school GPA

adequately predicts success in accounting courses, and other investigators (Bellico 1972, Cohn 1972, Ingram and Peterson 1987, Borde 1998) report that college GPA is a valid predictor of success in economics courses. Pre-university exposure to business-related courses is reported to have no effect or a negative effect on student performance in introductory businessrelated courses at the university level. Baldwin and Howe (1982) report that students who studied accounting in high school performed as well in an introductory accounting course at the university level as students who had no prior exposure. Bellico (1972) found that prior enrollment in community college economics courses negatively affected student performance in economics courses at the university level. Simpson and Sumrall (1979) and Borde, Byrd, and Modani (1996) report similar findings in finance courses. Surpassing the effectiveness of cognitive and academic variables in their apparent ability to predict student success in introductory business-related courses are the demographic and affective variables of gender and motivation. The effect of gender on success in business-related courses is significant (Siegfried 1979, Heath 1989), and seems to become more pronounced in courses in which analytic exercises become more advanced (Anderson et al. 1994). Gender also seems related to attrition. Male students seem more likely than female students to persist in economics courses (Hovrath et al. 1992).

12

Predictors of Student Achievement in Mathematics

The cognitive factors that have been most widely considered as potential predictors of college mathematics achievement are Scholastic Aptitude Test (SAT) and American College Testing Program (ACT) scores. Troutman (1978) and Bridgeman (1982) both found significant relationships between SAT Math scores and student achievement in college algebra and finite mathematics, respectively. Gussett (1974) found strong correlations between SAT Total (Math and Verbal combined) scores and grades in a suite of freshman-level mathematics courses. Likewise, Kohler (1973) found that ACT Math and Composite (Math and English combined) scores were significant predictors of grades in college algebra. Edge and Friedberg (1984) found that ACT Math, English, and

Composite scores were significant predictors of grades in calculus. And House (1995) found that the ACT Composite score was a significant predictor of grade in a variety of introductory college mathematics courses. Other researchers

found that combining admissions test scores with high school performance data successfully predicted grades in a variety of college math courses. Richards et

al. (1966) found that high school grades were good predictors of college math grades, especially when combined with ACT scores. Noble and Sawyer (1989) showed similar results in six college math courses using a combination of ACT Composite scores and high school GPAs. Keeley et al. (1994) found that

13

combining admissions test scores with high school rank predicted grades in numerous lower- and upper-division math courses. Troutman (1978) also reports that high school rank and grades in mathematics are good predictors of success in college mathematics. While many researchers report that standardized test scores and high school grades are effective predictors of success in college mathematics, some researchers report contrary findings. For example, Haase and Caffrey (1983a, 1983b) found that high school grades were almost useless as predictors of grades in introductory mathematics courses, and that SAT and ACT scores did not predict overall scholastic achievement in community college. Yellott (1981) reported that neither the ACT nor results from the Mathematical Association of America Placement testing Program tests predicted success in university level developmental mathematics courses. Despite these contrary findings, the

majority of researchers seem to agree that standardized test scores and high school grades are effective predictors of success in university-level mathematics courses. Many studies examined the utility of nationally administered aptitude tests, but some studies investigated the utility of locally administered subject- or course-specific exams. Crooks (1980), Bone (1981), Helmick (1983), and Shultz and Austin (1987) all found that subject-specific placement exams written and administered by the same institutions that taught the math courses in their respective studies were the best predictors of student performance in those courses. Crooks (1980) also showed that high school rank and GPA, as well as

14

scores from standardized achievement tests were strong and comparable predictors of college math grades. In addition to cognitive and quantitative factors, noncognitive factors have been used successfully to predict grades in college mathematics. Meece et al. (1982) found a relationship between student motivation, academic self-concept (a student’s personal opinion toward her or his academic skills), and achievement in introductory math courses, and an associated relationship between initial achievement and downstream persistence in more advanced math courses. Academic self-concept was shown to be a strong predictor of persistence in undergraduate math programs (House 1992) and final grades in math courses (Wilhite 1990, Gerardi 1990, Astin 1993, and House 1995). Interestingly, House (1995) found that academic self-concept specific to mathematical ability was a stronger predictor of final grade than any cognitive factors (including ACT scores) measured, and that this academic self-concept was a stronger predictor of final grade for females than for males. Factors that were considered but not found to be significant predictors of achievement in introductory math courses include the number of years of high school mathematics taken and student self-confidence in overall intellectual ability (House 1995).

15

Predictors of Student Achievement in Computer Science

Most studies investigating predictors of performance in college-level introductory computer science and/or computer programming courses report that aptitude in mathematics, measured by grades in high school mathematics courses or performance on institution or course entrance examinations, is the most salient predictor of success (Alspaugh 1972, Peterson and Howe 1979, Kurtz 1980, Fowler and Glorfeld 1981, Hostetler 1983, Konvalina et al.. 1983, Scymczuk and Ferichs 1985, Oman 1986, Cantwell Wilson 2002, Fan et al.. 1998). Fan et al.. 1998 report that math proficiency is a more accurate predictor of success in college computer science courses than standardized college entrance exam scores. Some studies report that factors related to student interaction with the curricular materials are relevant to student success. Violet (1997) reports that student effort predicts achievement. McGill et al. (1997) report a significant

relationship between success and the number of hours per week students engage in practical work (i.e. programming, homework assignments), but no relationship between success and the time invested in studying theory. Other studies report that prior experience with computers is an important predictor of success in collegiate computer science courses. McGill et al. (1997) report that students with previous programming experience are less likely to droup out of computer science courses than students with no previous programming experience. Taylor and Moundfield (1991) found that having a

16

structured computer-programming course in high school is a significant predictor of success in undergraduate computer science courses. Taylor and Moundfield (1994) report that any type of pre-college exposure to computers improves the likelihood of success for females. However, predicting success for males seems more complicated. General experience with computers did not differentiae

successful from unsuccessful males; the only factor that differentiated males was participation in a pre-college computer course that involved computer programming. Cantwell Wilson (2002) reports that formal classroom exposure to computer programming is a positive predictor of success, while computer game playing is a negative predictor of success. Clarke and Chambers (1989) report that experience with computers, including computer gaming, is the best predictor of both grades and persistence in tertiary computer science courses. Numerous affective predictors of student performance are also

documented in the literature. McGill et al. (1997) report that students’ level of confidence in their ability to pass the course, and perception of the importance of the need to seek tutorial assistance, predict student success. Cantwell Wilson (2002) reports that student comfort level, even more than math background, is the best predictor of success in introductory level computer science classes. Numerous studies report that students who attribute their successes in computer science to skill and their failures to bad luck are generally successful, while students who attribute their successes in computer science to good luck and their failures to lack of experience or ability are generally unsuccessful (Clark and

17

Chambers 1989, Bernstein 1991, Howell 1993, Moses 1993, Pearl et al. 1990, Cantwell Wilson 2002).

Predictors of Student Achievement in Physics

Several studies report simple correlations between good high school grades, academic preparation, and success in introductory physics courses in college. Gifford and Harpole (1986), Hart and Cottle (1993), and Alters (1995) all report that students who had good grades in high school mathematics and had taken physics in high school performed well in introductory physics courses in college. However, Champagne and Klopfer (1982) and Halloun and Hestenes (1985), found that these correlations did little to explain the actual cause of strong performance and deep conceptual understanding in college physics. Both Champagne and Klopfer (1982) and Halloun and Hestenes (1985) found that student preconceptions of physics concepts affected student success in college physics significantly, and that performance on specialized conceptual tests to identify these preconceptions was a better predictor of college physics grades than either high school grades or academic preparation was. Sadler and Tai (2001) report that rigorous preparation, including calculus and two years of physics, in high school predicts high grades in college physics. Additionally, Sadler and Tai (2001) and Tai and Sadler (2001) report that

18

variables not related to student preparation, but are instead related to curricular design and pedagogy were salient predictors of student success. Curricula that moved slower and addressed fewer concepts in more depth, and classroom cultures that deemphasized reliance on the text and quantitative problem solving were more successful at helping students achieve higher grades and deeper understandings of course materials. Sadler and Tai (2001) report additional interesting trends that are relevant to student success in college physics. For example, they report that college physics students perform better in classes that are taught by an instructor that is the same gender as the student. Sadler and Tai (2001) and Tai and Sadler (2001) report that, when other factors are controlled for, females perform better in algebra-based college physics, but males perform better in calculus-based classes. Sadler and Tai (2001) also report that Asian and white students, students from affluent communities, and students whose parents had advanced educations, tend to perform better in college physics than their peers who are black and Hispanic, who are raised in socioeconomically disadvantaged communities, and whose parents are less well educated. Similarly, Neushatz and McFarling (1999) report that students from socioeconomically disadvantaged communities are less likely to take physics in high school. Gender, race,

socioeconomic status, parental education, and educational achievement are factors that are commonly correlated, suggesting a common underlying cause. These factors and their potential underlying causes have received little attention

19

in formal research investigations on student success in college business, mathematics, and science courses.

Predictors of Student Achievement in Chemistry

Various measures of student cognitive ability have been used as predictors of achievement in undergraduate college chemistry. Numerous

studies have shown college admissions test scores to be significant predictors of achievement (Craney and Armstrong 1985, Ozsogomonyan and Loftus 1979, Andrews and Andrews 1979, Pederson 1975, and Reiner 1971; although see House 1995). Other studies have found that advanced logico-mathematical

reasoning skills are important for success in freshman chemistry (BouJaoude and Giuliano1994, Niaz and Robinson 1992, Chandran et al. 1987, Demko et al. 1985, Good 1983, Howe and Durr 1982), although such skills may only account for between 21% (Albanese et al. 1976) to 15% of the variance in student grades, leaving 85% of the variance to other variables (Good 1983). In addition to standardized reasoning tests, locally developed placement tests have been found to be effective predictors of student success in freshman chemistry. Wagner et al. (2002) report that performance on their “Student PreSemester Assessment” test predicted the pass/fail status of 41% of their general chemistry students, while the College Board’s Scholastic Aptitude Test only predicted pass/fail status at 17%. Tests of prior content knowledge and/or

20

academic experience in chemistry have also been shown to be strong predictors of success in freshman chemistry (Yu 1999, BouJaoude and Giuliano1994, Chandran et al. 1987, Coley 1973). Measures of noncognitive student variables such as initial attitudes toward chemistry, academic self-esteem (particularly self-rating of mathematical ability) and achievement expectancy have been reported to be better predictors of student success in college chemistry at a large public university than are cognitive variables (House 1995). However, Ferarri and Parker (1992) report that high school achievement (measured by grade point average) is a better predictor of achievement in college chemistry than initial student attitudes such as global (both academic and non-academic) self-efficacy. Additionally, Okebukola (1987) reports that student attitude, as well as a classroom climate that emphasizes student participation in laboratory activities, were the best predictors of student success in chemistry in 37 secondary schools.

Predictors of Student Achievement in Biology

Numerous studies have shown that logico-mathematical skills are strong predictors of advanced performance in secondary and tertiary biology. Numerical and analytical skills measured by the quantitative section of the College Board’s Scholastic Aptitude Test are reported to be predictors of student achievement (Helseth et al. 1981, Yeany et al. 1981). Arithmetical skills (Detloff

21

1982) and general mathematical skills (Marsh and Anderson 1985, Biermann and Sarinsky 1989) are also valid predictors of success in freshman biology. More general reasoning skills, such as those measured by Piagetian or neo-Piagetian logic questions (Piaget 1966), have also been demonstrated to be strong predictors of student success in college biology (Bullock et al. 1976, Dettloff 1982, Helseth et al. 1981). Davidson and Haffey (1979) suggest that a student’s intelligence quotient (IQ) is the best predictor of her or his success in high school biology. In addition to logic and reasoning skills, background knowledge in biological concepts also seems important to success in college life science courses. Pretests that measure students’ biological background knowledge have been shown to be useful predictors of understanding advanced biological concepts such as evolutionary theory (Lawson 1983), student success in college biology courses (Hooper 1968), and success in programs designed to prepare students for advanced study in various health professions (Carmichael 1986). Interestingly, a variety of studies have demonstrated that verbal skills related to reading and comprehension are the most salient predictors of success in college biology. Several studies (Emmeluth 1979, Detloff 1982) report the usefulness of the Nelson-Denny Reading Test (NDRT) as a predictor. The

NDRT is a timed test that measures vocabulary development, comprehension, and reading rate. It is widely used as a reading placement test in American colleges and universities. Emmeluth (1979) reports that the NDRT

comprehension score is a more valid predictor for women, and that the NDRT

22

vocabulary score is a more valid predictor for men. However, the value of the NDRT as a predictor is not without controversy. Gudan (1983) reports that the NDRT did not predict grades in two different introductory biology courses. Like the NDRT, the verbal section of the College Board’s Scholastic Aptitude Test (SAT-V) has demonstrated value as a predictor of grade in biology. Nist et al. (1995), Marsh and Anderson (1985), Yeany et al. (1981), and Szabo (1969) all report that students’ SAT-V scores are valid predictors of success in freshman biology. Prior academic performance has also been shown to be a significant predictor of success in introductory biology. Szabo (1969) reports that

performance in high school science predicts achievement in college life science courses. Other studies suggest that high school grade point average and/or rank are strong predictors of success (Hooper 1968, Emmeluth 1979, Yeany et al. 1981, Marsh and Anderson 1985, Carmichael 1986). Finally, student perceptions seem to be important components of success in college biology. Pridmore and Halyard (1980) report that student outcomes on portions of the Academic Motivations Inventory, when coupled with other quantifications of student aptitude such as grade point average or scores on the Scholastic Aptitude Test, can be used to predict student academic success in biology. Nist et al. 1995 found that student self-perception of examination

performance was also a valid predictor of final grade. The authors suggest that accurate self-evaluation is a metacognitive talent that is well-developed in successful students.

23

Summary of Factors That Predict Student Success in Introductory Business, Mathematics, and Science Courses Three variables seem to be general predictors of success in freshman business, mathematics, and science courses. students’ quantitative and analytical skills. One variable is cognitive:

Standardized tests such as the

mathematics sections of the ACT or the SAT, and local tests that contain neoPiagetian questions and questions focused on course-specific logical skills, provide data relevant to this variable. Data from such tests are easily generated or acquired and readily interpretable. A second general variable that predicts success is affective, and relates mostly to students’ academic self-esteem. Measures of academic self-esteem are less not as widely available as measures of mathematical skills, but numerous instruments are available for measuring this variable. The results

these instruments yield, however, might not be as easily interpretable as results from a mathematics test. Mathematics tests typically contain questions that have correct or incorrect answers, whereas instruments that measure student affection typically generate graded responses. Furthermore, student affection can vary from course to course, teacher to teacher, even day to day. Still, reliable and valid methods of measuring student affection, including academic self-esteem, exist.

24

The third nearly universal variable that predicts student success in freshman business, mathematics, and science courses is high school grade point average. Grade point average is neither a cognitive nor affective variable; it is neither a measure of aptitude nor state of mind. Instead, it is a holistic measure of performance. Both cognitive and affective states influence it. Similar to data on mathematical skills, data on students’ grade point averages are widely available. But the interpretation of students’ GPAs is potentially more

challenging, since GPAs are a composite measure of a student’s overall high school experience. In addition to these nearly universal predictors, several subject-specific predictors are documented by multiple researchers. The value of attribution to success in computer science is well demonstrated, and might be an affective characteristic related to academic self-esteem. Hands-on experience with

computers, both before college and in class also seems to enhance a student’s chances of success in this field. Experience also seems to be valuable for success in other fields of science. In physics, prior experience with complex physical concepts and

theories seems at least as valuable as hands-on experience with physical phenomena. The same has been shown to be true in biology. This could be true because most university-level science courses require students to grasp concepts associated with post-formal operational reasoning, and time, intellectual maturity, and experience are all required for post-formal concept construction (Lawson et al. 2000a, Lawson et al. 2000b). If this were true, then one would

25

expect that experience also play an important role for success in chemistry. However, the effect of experience on success in chemistry has not yet been thoroughly investigated. The finding that academic experience can have a negative effect on student success in economics and finance courses is interesting and not well explained. It is possible that the concepts that are introduced in community

college economics and finance courses are different than – or even in conflict with – concepts introduced in economics and finance courses at the university level. It is also possible that university level economics and finance courses are more tailored for students who intend to continue their educations in these fields, whereas courses at the community college level are tailored for people who do not plan to continue. Educational content of courses and pedagogy could

therefore be different at the two kinds of institutions, and students with experience at the community college level might have a different perception of requirements for success in these courses than students at the university level. Equally as interesting is the finding that verbal skills are valid predictors of success in biology courses. This phenomenon is also poorly addressed in the research literature, but worthy of pursuit. It is possible that verbal skills help students understand and articulate the critical qualitative arguments that accompany quantitative concepts that constitute deep understanding of biological theories. Such qualitative arguments could be less common and/or important in other natural sciences, including chemistry and physics.

26

Interpreting Results of Predictive Studies in Business, Mathematics, and Science Education Although most studies to date have described what seem to be legitimate predictors of success in introductory college business, mathematics, and science courses, these results must be interpreted with caution. All published studies reviewed for this manuscript aspired to find predictors of success in business, mathematics, and science courses, and all studies successfully found them. But few of the studies were actually experimental; most protocols involved post hoc comparisons of student grades with other variables. Thus, the results of most studies are correlative, not causal. Factors that ostensibly cause student

success in business, mathematics, and science courses – i.e. factors that cause the reported correlations to exist – could be different than those reported in the predictive studies. For example, numerous authors report that scores on exams that measure quantitative and analytical skills correlate strongly with final grades in business, mathematics, and science courses. But what causes a student to receive a high (or, alternatively, low) grade on an exam that putatively measures quantitative and analytical skills? One obvious possibility is student aptitude. But other

possibilities might include the career choices or education level of the student’s parents, who could be mentoring the student in this area of achievement. Or

27

perhaps high scores on exams that measure quantitative and analytical skills were driven buy access to educational resources and opportunities, such as attendance at a summer “math camp,” or participation in an extracurricular test preparation course. Access to such resources might ultimately be determined by the students’ socioeconomic circumstances. These circumstances could be a causal factor driving the student’s exam score. Standardized test scores also seem to have different predictive value for women and men. Brush (1991) reports if a woman and man have the same SAT scores entering college, the woman is likely to achieve higher grades in college. Said differently, women who have lower SAT scores will perform comparably to men who have higher scores. Behnke (1989) reports that, at the Massachusetts Institute of Technology, women who scored 20-25 points lower on the math section of the SAT achieved grade point averages comparable to their male peers in science, math, and engineering courses. Related to academic self-esteem, student attitude also influences performance in science classrooms. Students with more positive attitudes

toward science tend to do better in science courses (Weinburgh 2000, Weinburgh 1994, Oliver and Simpson 1988, Kaballa and Crowley 1986, Willson 1983, Gardner 1975, Ormerod and Duckworth 1975). Females typically have more negative attitudes toward science than do their male peers (American Association of University Women 1992). Another example of a result that might be challenging to interpret is the finding that a student’s academic self-esteem correlates strongly with her or his

28

final grade. Students with high academic self-esteem tend to do well in business, mathematics, and science courses. But what causes a student to have high academic self-esteem? Intrinsic confidence in one’s own intellectual abilities is an obvious possibility. But intrinsic self-confidence in college science is not

equivalent among females and males. Females tend to perceive their cognitive styles as imaginative and intuitive, and inconsistent with the rote, serious, and competitive culture of most college science classrooms. Women also raise their hands and manipulate laboratory equipment less frequently, and prefer to work in groups more frequently, than do men (Tobin and Garnett 1987). Forms of

engagement that are preferred by females might be discouraged in college science classrooms and labs and contribute to the discomfort of women students. Using engagement techniques that are discouraged in the classroom, or having the perception of being out of place in the classroom could affect a female student’s level of comfort (Cantwell Wilson 2002) or belief in her ability to succeed (Brush 1991, Bar-Haïm and Wilkes 1989, Blenkly et al. 1986). In addition to intrinsic factors, extrinsic factors might also affect a student’s academic self-esteem. Again using gender as an exemplar, women and men often respond to pedagogical styles and classroom cultures differently. Many women prefer and are more comfortable in classrooms where deliberation and collaboration are more common than memorization and competitiveness (Tobias 1990). Most introductory business, mathematics, and science courses have

classroom cultures that alienate instead of encourage female students (Constantanople et al. 1988, Hall and Sandler 1982). In science classrooms,

29

men are engaged more by teachers (Tobin and Garnett 1987), and male role models – including the professor and teaching assistants – are more common than female role models (Brush 1991, Hall and Sandler 1982). Such pedagogical and cultural inequities can have significant negative effects on the academic selfesteem of female business, mathematics, and science students.

Methods

To determine who receives DFWs in gateway business, math, and science courses at NAU and to investigate why these students receive DFWs in these courses, three types of data were collected and analyzed. General student background data such as demographic, standardized test score, and grade information were obtained from NAU’s Office of Planning and Institutional Research. Publicly available demographic and performance statistics about high schools from which in-state students originated were collected from the U.S. Department of Education, the Arizona Department of Education, and a nonprofit K-12 education advocacy organization named GreatSchools, Inc. Data about student motivations and social habits were collected by surveying a large group of students enrolled in gateway courses of interest. collection is described in detail below. All information that could be used to personally identify study participants was removed before data were analyzed. The study was performed in Each method of data

30

compliance with policies and regulations regarding the use of human subjects in research, and under the supervision of NAU’s Institutional Review Board. Unless otherwise noted, all statistics were calculated using JMP® IN
P

Version 4 Release 4.0.4 (SAS Institute Copyright © 2001).

Institutional Records and Public Data

NAU’s Office of Planning and Institutional Research (OPIR) collects data on an ongoing basis on a variety of student demographic and academic attributes. These data are warehoused and made available on request for

institutional research. Student data (n = 23255) from the 15 gateway courses listed in Table 1 and taught in regular semesters from Fall 1997 through Fall 2001 were requisitioned from this source. Data that were obtained include age, gender, ethnicity, high school name, high school grade point average, high school class rank, college hours completed, cumulative college grade point average, current semester hours enrolled, current semester grade point average, American College Testing (ACT) score, Scholastic Aptitude Test (SAT) score (a composite of critical reading, math, and writing scores), and major. Data on student final grade in each course were also obtained.

31

Prefix ACC 255 ACC 256 BA 201 BIO 100 BIO 181 BIO 182 CHM 151 CHM 152 CIS 120 ENV 101 GLG 100 MAT 125 MAT 137 PHY 111 PHY 112

Course Name Accounting Principals Financial Accounting Principals Managerial Quantitative Methods (Business) Principals of Biology Unity of Life I Unity of Life II General Chemistry I General Chemistry II Introduction to Computer Information Systems Environmental Science Introductory Geology Pre Calculus Calculus II General Physics I General Physics II

n 2198 1154 1151 1881 1771 558 2373 1129 4114 888 1714 2009 726 713 394

Table 1: Courses from which institutional data were collected

The name of the high school each in-state student attended was also obtained from the OPIR. A list of the high schools that were included in the study (n = 244) is shown in Appendix A. Public records for each school were searched, and demographic information was collected. Information collected includes the average Arizona’s Instrument to Measure Standards (AIMS) reading and math scores, average Stanford 9 (SAT-9) reading and math scores, the average ACT and combined SAT scores, and the percent of the student body that qualifies for the federal free or reduced price lunch program. This statistic is commonly used as a measure of socioeconomic status of high schools and the communities and families they serve.

ABC and DFW Rates in Gateway Courses

32

Student data were placed in either the ABC or the DFW group based on the final grade students received in each course. Since NAU sometimes awards grades that are more descriptive than traditional letter grades, students who received nontraditional grades were also assigned to one of the two categories. The categories to which each nontraditional grade was assigned are listed in Table 2. Nontraditional grades that could not logically be assigned to either the ABC or the DFW category were categorized as “Not counted.” Grades

categorized as “Not counted” included audits, incompletes, and grades of “P” in pass/fail courses. Data categorized as “Not counted” were excluded from the analysis.

Grade A A# AU B B# C C# D D* F F* I P W

Description Earned A Earned A, Repeat Audit Earned B Earned B, Repeat Earned C Earned C, Repeat Earned D Repeat Replaced Earned F Repeat Replaced Incomplete Pass Only Withdrawal

Category ABC ABC Not counted ABC ABC ABC ABC DFW DFW DFW DFW Not counted Not counted DFW

Table 2: Grades reported in gateway courses of interest with their designation in grades by course analysis

ABC and DFW rates in 13 gateway business, math, and science courses at NAU were calculated. Data from BIO 181 and BIO 182 were incomplete and therefore excluded from the analysis. Rates for fall and spring semesters were

33

calculated separately and then averaged. To determine if there was a difference in ABC and DFW rates between fall and spring semesters, Student’s t-tests were performed.

Characterizing ABC and DFW Students

Two sources of data were used to characterize ABC and DFW students. The primary source of data was a student survey administered in seven gateway courses in 2002. This 26-item multiple response survey queried gateway students on their demographics, academic habits, motivations, and attitudes related to college and gateway courses. Student survey data were supplemented by student demographic and academic qualification data supplied by NAU’s OPIR. These institutional data were used to elucidate or confirm missing, ambiguous, or sensitive results derived from the student survey. Because most variables in this study are categorical, contingency analyses were usually performed to determine if nonrandom relationships among variables exist. Contingency analyses traditionally yield either a Pearson chisquare (X2) or a likelihood ratio (G2) test statistic. Under normal circumstances, both these statistics are equivalent and can be interpreted as such. Under some circumstances, such as when sample sizes (n) are unusually high or when some µi (means of cells) are less than 0.5, the X2 and G2 statistics diverge. In these circumstances, G2 is a usually more conservative measure of the effect size than

34

is X2 (Agresti 2002). categorical analysis.

For this reason, G2 is the statistic reported for each

In some cases, the sample size in each cell of the contingency table (c) is less than 5. In these circumstances, contingency analyses produce results that are suspect (Agresti 2002). When results in this study were suspect for this reason, sparse categories were collapsed into categories that were less descriptive but that provided larger sample sizes per cell. When categories were collapsed, analyses were re-run. Results that were consistent with the results of the original (suspect) analysis were reported with a cautionary note. Results that were not consistent with the results of the original (suspect) analysis were not reported, and the original (suspect) analysis was not reported as significant.

Student Survey

A 26-item survey (Appendix B) designed to assess student attitude toward the University, gateway classes at the University, and personal academic habits was written and administered to 719 students in seven gateway classes (ACC256, BA201, CIS120, ENV101, MAT125, PHY111, and PHY112) during the Spring 02 and Fall 02 semesters. Student participation was voluntary. The survey was developed in October 2001 by representatives from NAU’s Science and Math Learning Center, the College of Social and Behavioral Science, and Office of Student Life. Questions were mostly derived from

administrative officials, course instructors, and education researchers at NAU.

35

The survey contained questions regarding student demographics, academic performance, preparation, study habits, learning styles, goals, obstacles, motivations, and perceptions toward the class and the University. Response options on the survey were multiple choice. Multiple-choice answer options were derived mostly from a free-response pilot version of the survey administered to 124 BIO100 students in November 2001. The most

common responses from that version were incorporated in the response options of the multiple choice survey that was administered and that provided data for this report. Each participant provided her or his student identification number on the survey response sheet. Student identifiers were used to obtain students’ final grades in surveyed courses, and to determine if any student completed the survey in more than one course. Final grades were obtained from NAU’s OPIR. If a student completed the survey in more than one course, data from only one of the courses were used in the analysis to prevent pseudoreplication. If a student completed the survey more than once because she or he was enrolled concurrently in two or more courses, data from only one of the courses were used. The course from which data were obtained was randomly chosen. If a student completed the survey in consecutive semesters, data from only the first (i.e. the Spring) semester were used. When final grades were obtained and pseudoreplicates were eliminated, student identifiers were removed. survey results were ultimately made anonymous. Thus,

36

After survey data were collected, demographic data were used to characterize the student body in gateway courses, and descriptive statistics were performed. Data were then divided into ABC and DFW groups, and analyzed to determine what, if any, differences exist between the two groups. Student’s ttests were used to compare continuous quantitative data such as age, standardized test score, and grade point average. Log-likelihood tests were used to compare nominal and ordinal categorical qualitative data such as gender, ethnicity, and level of academic achievement. Results in numerous areas of analysis were produced. Demographic data were used to describe student perception of course, student academic habits, effect of course on student, and student perception of college life and NAU. ABC and DFW data were used to investigate hypotheses regarding student success in courses of interest. These hypotheses were derived from two sources. One source was the primary literature, which proposes a variety of causes of student success in gateway courses. The other source was NAU instructors who teach gateway business, math, and science courses. These instructors provided numerous ideas about determinants of student success in their courses. There was a surprising consistency of opinion among gateway course instructors about why students do or do not succeed. Hypotheses that were offered and investigated included ethnicity, gender, student opinion of course, student perception of academic status in the course, student academic qualifications, impact of course on student goals and interests, attendance, and study habits.

37

Characterizing Students’ Educational and Socioeconomic Contexts

To determine what, if any, effects the educational and socioeconomic contexts from which students originated effected student success in gateway courses, students were grouped together by high school of origin, and the final grades each student received in the first gateway business, math, or science course in which they enrolled at NAU were compiled. If students were enrolled in two courses concurrently, data from only one of these courses were used in the analysis to prevent pseudoreplication. In these instances, the course from which data were used was randomly chosen; data from other courses were excluded from the analysis. The rate at which students from each high school received a D, F, or W in these courses was then calculated, yielding a single DFW rate for the group of students who attended each high school. To determine if demographic characteristics of high schools and/or neighborhoods of student origin correlated with student achievement in gateway courses, Pearson product-moment correlations were performed on the DFW rates, average standardized test scores, and rates of reduced cost lunches from each high school.

38

Characterizing Gateway Classrooms and Courses

To investigate the hypothesis that characteristics of the course affect student success and failure rates, observations were made in one randomly chosen section of each of the gateway courses of interest. Courses were characterized with the Reformed Teaching Observation Protocol (RTOP) (Piburn et al. 2000, Sawada et al. 2002). The RTOP (Appendix D) consists of 25 statements about components of instructional practice such as lesson design and implementation, course content (both propositional and procedural knowledge), and classroom culture (both communicative interactions and student-instructor relationships). Each of the 25 statements is scored on a 0–4 ‘‘Never Occurred’’ to ‘‘Very Descriptive’’ scale. Thus, the RTOP allows observers to rate instruction on a 0–100 scale. This RTOP score describes the extent to which reformed instructional practices (Alexander and Murphy 1999; National Council for the Teaching of Mathematics 1989, 1991, 1995, 2000; National Academy of Sciences, National Research Council 1996, 2000; American Association for the Advancement of Science 1989) are used. Each course in the study was visited once during two semesters during the span of the study (i.e. in the Fall 1997 through Fall 2001 semesters). RTOP scores were calculated for courses of interest not to describe the instructional practices employed in each course, but to describe the range and variability of instructional practices employed in all gateway courses. To this end,

39

descriptive statistics on course RTOP scores and of the scores of the subcategories within the RTOP were generated. Prior research has explored the relationship between the degree of instructional reform and student achievement. These studies have found that instructional reforms, as reflected by RTOP score, have had a positive effect on student achievement in college science and mathematics courses (Falconer et al. 2001, Lawson et al. 2002). Because these results were obtained in gateway mathematics and science courses at other universities, it is reasonable to predict that similar effects might be seen in courses of interest at NAU. To determine if there was a relationship between instructional strategies used in gateway courses at NAU and student success, Pearson product-moment correlations were performed on course RTOP scores and course ABC rates.

Development of Predictive Model

To develop predictive models for identifying students who might be at risk of receiving a D, F, or W in a course of interest, a stepwise multiple logistic regression was used. Stepwise regression is a statistical technique used to “Best”

identify a “best” set of predictors from among a variety of variables.

describes a set of variables that is maximally parsimonious and satisfactorily predictive for the requirements of the research (Sokal and Rohlf 1995). In this study, coefficients of determination (R2) were calculated for all variables that were hypothetically related to student success and for which data were available.

40

The model was then generated by starting with the variable with the highest R2 and adding and eliminating other variables to the model until the best model was obtained. Multiple logistic regression is a statistical technique that employs numerous X variables to predict a single, binomial Y outcome (in this case, Y = membership in either the ABC or DFW group). To fit a single regression line to the logit-transformed data, the maximum likelihood method was used. Because one or more of the assumptions generally associated with Model I regressions (no sampling error, Y is a linear function of X, independence, normality, and homoscedasticity [i.e. equal variance around the regression line]) were likely violated, a Model II regression for predicting ABC or DFW status was performed (Sokal and Rohlf 1995). Two types of data – “intrinsic” data describing the academic habits and achievements of individual students, and “extrinsic” data describing the demographics and average academic performance of students’ high schools or students from those high schools taking gateway courses at NAU at the time the study was conducted – were collected, two separate models were used in the regression analyses. The X variables that were considered in both models are listed in Tables 3 and 4.

41

Age Cumulative college credits earned Cumulative college grade point average Current semester credits enrolled Current semester grade point average Ethnicity Gender High school grade point average High school rank Table 3: X variables considered in regression model for “intrinsic” student data. Variables are listed alphabetically.

ACT score Age Average high school AIMS math score Average high school AIMS reading score Average high school grade point average Average high school rank Average high school Stanford 9 math score Average high school Stanford 9 reading score Cumulative college credits earned Cumulative college grade point average Current semester credits enrolled Current semester grade point average Ethnic proportions Percent of females SAT (combined) score Socioeconomic status Table 4: X variables considered in regression model for “extrinsic” student data. Variables are listed alphabetically.

The criteria used for developing predictive models were parsimony and utility. Models with fewer predictor variables and predictor variables that were universally available (e.g. high school grade point average, which is available for all gateway students, rather than ACT score, which is only available for a subset of gateway students) were preferred to models with many or sparsely distributed variables. Furthermore, models that had high predictive values and could be used for all (not just a subset of) gateway students were preferred.

42

Results

Data on student success rates in gateway business, math, and science courses at NAU are presented below. Statistics on course-oriented variables are presented first. Next, data on student-oriented variables are presented. Finally, several models designed to predict students’ ABC or DFW status are described. Data are presented this way to separate external/contextual factors from internal/personal factors. It is hoped that organizing the data in this fashion will illustrate interesting trends and allow consumers of this information better design appropriate and effective interventions to address specific concerns. Statistical significance was determined when p ≤ 0.05. Confidence was described as “approaching significance” when 0.05 ≤ p ≤ 0.10. In instances when this occurred, statistical results were reported. When p ≥ 0.10, results were considered to be non-significant, and they were not reported. Graphs are provided to help illustrate interesting trends and significant findings. When significant differences between genders or among ethnic groups are present, graphs to illustrate these between and among group differences are provided.

43

Course-Oriented ABC and DFW Statistics

Because a significant percentage of grades were not reported at the time of data collection for BIO 181 in the Fall 1997 through Spring 2001 semesters (average number of grades not reported = 46%) and BIO 182 in the Fall 2000 through Spring 2001 semesters (average number of grades not reported = 49%), valid ABC and DFW statistics for these courses could not be accurately calculated. Thus, BIO 181 and BIO 182 data were excluded for most analyses in this portion of the study, although data from these courses were included in other portions of the study because these other types of data do not rely on a majority of final grades being reported to ensure their validity. A significant number of grades were similarly not reported for CIS 120 from Fall 1997 through the Spring 1999 semesters (average number of grades not reported = 49%). However, grade data were complete for the Fall 1999 through Spring 2001 semesters. These data were used to calculate this class’ ABC and DFW statistics.

ABC and DFW Rates in Gateway Courses

Average ABC and DFW rates for 13 gateway business, math, and science courses taught in the 1997 through Fall 2001 semesters at NAU are reported in Table 5. The average ABC rate was 75%, and the average DFW rate was 25%. There was no significant difference between ABC and DFW rates in the Fall (n = 13, t = -0.716, p = 0.242) and Spring semesters (n = 13, t = 0.181, p = 0.571).
44

While most courses’ ABC and DFW rates fall within one standard deviation of the mean, two courses fall outside of that distribution. PHY 112 has a comparatively high ABC rate (and thus low DFW rate), and MAT 125 has a comparatively low ABC rate (and thus high DFW rate).

ACC255 ACC256 BA201 BIO100 CHM151 CHM152 CIS120 ENV101 GLG100 MAT125 MAT137 PHY111 PHY112 Mean SD

Spring ABC Rate 66% 70% 69% 79% 68% 78% 82% 81% 84% 60% 67% 83% 84% 75% 8%

Fall ABC Rate 69% 69% 73% 81% 73% 74% 85% 78% 82% 60% 72% 82% 83% 75% 7%

Spring DFW Rate 34% 31% 32% 22% 33% 21% 19% 21% 16% 40% 33% 17% 16% 26% 8%

Fall DFW Rate 31% 32% 27% 19% 27% 25% 16% 22% 18% 40% 28% 17% 23% 25% 7%

Average ABC Rate 67% 69% 71% 80% 70% 76% 83% 79% 83% 60% 69% 83% 84% 75% 8%

Average DFW Rate 33% 31% 30% 20% 30% 23% 17% 21% 17% 40% 30% 17% 16% 25% 8%

Table 5: Grades reported in gateway courses of interest. Some totals may not equal 100% because of excluded data, missing data, reporting errors, and/or rounding errors).

Teaching Styles Used in Gateway Courses

A description of the teaching technique used in each course of interest, measured by the average RTOP score, is shown in Table 6. The range of scores was 24.0 – 77.0, and the mean score was 51.6 (SD ± 15.7), suggesting that most gateway business, math, and science courses at NAU are taught with relatively traditional, didactic methods.

45

Prefix
ACC 255 ACC 256 BA 201 BIO 100 BIO 181 BIO 182 CHM 151 CHM 152 CIS 120 ENV 101 GLG 100 MAT 125 MAT 137 PHY 111 PHY 112

Course Name
Accounting Principals Financial Accounting Principals Managerial Quantitative Methods (Business) Principals of Biology Unity of Life I Unity of Life II General Chemistry I General Chemistry II Introduction to Computer Information Systems Environmental Science Introductory Geology Pre Calculus Calculus II General Physics I General Physics II

Average RTOP Score 44.5 31.0 52.0 51.5 24.0 52.0 28.0 41.0 61.5 68.0 63.0 57.0 56.0 68.0 77.0

Table 6: Description of teaching styles, measured by the RTOP, used in each course. Low scores suggest didactic techniques; high scores suggest reformed techniques.

While most courses had average RTOP scores that were within one standard deviation of the mean, the RTOP scores of several courses were outside of this range. The average scores of three courses (BIO 181, CHM 151, and ACC 256) were lower than one standard deviation, and the scores of three courses (ENV 111, PHY 111, and PHY 112) were above one standard deviation. A moderate correlation existed between each course’s RTOP score and its ABC rate (n = 13, r = 0.575, p = 0.040). This correlation is illustrated in Graph 1. Because the ABC rates for BIO 181 and BIO 182 were not available, these courses were excluded from this analysis.

46

90 85 80 Average ABC Rate 75 70 65 60 55 50 20 30 40 50 60 70 80 90 Average RTOP Score

Graph 1: Correlation between each course’s average RTOP score and average ABC rate (n = 13, r = 0.575, p = 0.040).

Student-Oriented ABC and DFW Statistics

General student survey results, as well as results analyzed by ethnicity and gender, are listed below. Results are grouped in seven major categories: student demographics, student perception of course, student academic habits, effect of class on student, student perception of college life and NAU, student opinion of course, and student awareness of academic status. The number and percent of students responding to each answer option are provided. The sum of counts for each option might not equal the total sample size because not all students responded to each question. The sum of percents for each question might not total 100% because of rounding or invalid student responses that could not be included in the total.

47

Twenty-three percent of students asked to participate in the survey did so. Although students in all groups of interest provided valid data, students who received an A in the course in which they took the survey were better represented, and students who received an F or W in the course in which they took the survey were more poorly represented, in the data set. students who received a W were included in the data set. Only three

This is primarily

because the survey was administered past the drop/add deadline each semester, and most students who received a W were not present when the survey was administered. Thus, the DFW statistics that rely exclusively on this survey data might be preferentially biased toward students who receive grades of D and F, but not W. Students of both genders were equally represented in the sample. Students of all ethnic groups are not equally represented, nor are the distribution of ethnicities in this data set representative of the distribution of ethnicities in the general U.S. population (U.S. Census Bureau 2000). A comparison of the

distribution of ethnicities in each population is below (reporting and rounding errors cause both columns to not total 100%) and in Graph 2.

% (in sample) 1) 2) 3) 4) 5) 6) African American Asian American Hispanic Native American White/Caucasian Other 2 2 8 5 79 3

% (in U.S. population) 13 4 14 1 69 10 hours 0 hours