Correlates of achievement with online and classroom-based MBL physics activities.

Students from five high schools participated in a 2 to 4-week microcomputer based laboratory (MBL) physics curriculum in two groups. One group completed the curriculum entirely online, and the other completed the same curriculum in a traditional classroom setting. Variables were collected to predict student success on a post-unit measure of physics ability through correlations. Traditional, literature-suggested, variables such as math ability, physics aptitude, and demographic information were combined with variables related to the mode of delivery of instruction. Several computer-related variables, such as students' access to a computer at home and how often students accessed the Internet were included in the models. Math ability and physics aptitude dominated the models regardless of the method of delivery of instruction. Computer-related variables were included in the model for the online group; however, they were not significant.

INTRODUCTION

The ability to predict the success or failure of a student in a class is a very powerful tool. If specific skills are highly correlated with success then every effort should be made to ensure that all students have these necessary skills. If a lack of certain skills correlates highly with failure, then students should receive remediation until their skills match those of the students deemed likely to succeed. Studies have been conducted using a wide variety of intellectual, socio-emotional, and background variables in an attempt to predict students' success or failure. These studies have been conducted in an effort to help teachers understand their students, to assist universities in selecting students to enter their institutions, and for advising students on classes necessary as prerequisites to afford these students their greatest opportunity for success. Adding to this genre of research, this study examined if there were any traditional or computer-related variables that could predict success in a computer-based physics unit.

DISCUSSION

Review of Literature

Much of the research pertaining to correlating success or failure with certain variables was completed in the 1970s and 1980s. This may have been partially in response to high profile reports that linked student success with socio-economic background. In a review of the current literature of the time, Margrain (1978) concluded that little variance could be explained after accounting for general intelligence. Much of the research produced mixed results, as different variables accounted for different portions of the variances in achievement. Another difficulty in identifying a consistent set of variables was the wide variety of indicators of success used in the different studies as the dependent variable. The inclusion of variables outside the realm of the school, such as socio-economic and other background variables, also did not consistently explain student variances in success.

However, the large amount of still unaccounted for variance in student performance suggests that the students might not be the sole arbiters of their success. Their teachers' ability, personality, bias, methods, and numerous variables associated with the institution attended must also be conconsidered [sic] for complete and accurate prediction of academic performance. (Margrain, 1978, p.121)

Colleges and universities have also consistently used predictor variables to find attributes of students that can be used for admission policies. The most typical variables used for admission are standardized test scores, such as the ACT or SAT, and high school performance. These two variables, however, also show a discrepancy as to how well they predict the success of a student at a university. The highest levels of correlation are formed only under certain conditions. The students need to stay in dorms on campus, their freshman class has to have a relatively small enrollment (i.e., 500 or less), and have above average standardized test scores that have a large standard deviation (Munday, 1970). More recently, an examination of these same variables, standardized test scores and high school performance, found that they correlated well with students' academic performance at institutes of higher learning, but did not hold any predictive value for students' interest or enjoyment of their studies (Harackiewicz, Barron, Tauer, & Elliot, 2002).

Many departments and colleges within a university also use correlations to attempt to predict students' success in their programs and to determine at what level they should be placed within the program. A study of engineering students revealed that the single best variable for predicting success was math achievement (Levin & Wyckoff, 1988). In an effort to determine why only 40% of males and 33% of females persisted in the natural sciences through graduation, achievement in mathematics was again the single best indicator (Adair, 1991). The combined variable of high school GPA (grade point average) and ACT score was the best predictor of achievement in a series of college English and math courses (Noble & Sawyer, 1989). This correlation was then used to suggest models for placing incoming students into the math and English curriculum at the appropriate level.

When completing correlations or predictive models, one of the confounding variables is gender. In some cases gender correlates with success (Okpala & Onocha, 1988), and in other instances it does not (DeBoer, 1985; McCammon, Golden, & Weunsch, 1988). One commonality in several studies, however, is that women can be more accurately correlated with success than men (McCammon et al., 1988; Munday, 1970). In their study, McCammon et al., found a correlation for all the students in a 2nd year physics course. When they controlled for gender, however, no factors predicted success for men, and the correlations for success for women were much higher than for the entire group.

Attempts have also been made to determine different variables for men and women that would correlate with success. In a study with a low population of females (females = 23, males = 96) predicting success in high school physics, Ignatz (1982) found that males preferred a divergent structure. This structure provides for a greater variety and quantity of 'right' answers. Females, on the other hand, preferred convergent structures. This convergent structure worked toward finding one correct answer. Studying characteristics of males and females who succeeded or failed in their first college science class, DeBoer (1985) found the characteristic of rashness, as defined by the Omnibus Personality Inventory, was positively correlated with success in men and negatively correlated with success in women. The same study showed women who considered themselves hard working and goal orientated were the most likely to succeed.

In a study that is somewhat indicative of those in this area, Edge and Friedberg (1984) correlated a variety of academic and biographical variables with achievement in a 1st year university calculus class. In this case, none of the demographic variables, including gender, were significant. The best indictors were an algebra pre-test score and high school rank. Edge and Friedberg (1984) thought that class rank could be more than an achievement measure, "It may in fact be the case that rank in class represents a measure of competitiveness as well as a characteristic of long-term emotional adjustment ..." (p. 140)

Since the introduction of computers in education, researchers have also studied what variables would correlate with successful computer use. In a survey of over 1,800 eighth through twelfth grade students, males responded as being more interested in computers, more confident in their computer ability, and less likely to feel that computers could have a negative impact on society than females (Collis, 1985). After completing a computer course, males had a more positive attitude, and females a less positive attitude, toward computers. The same study linked math and science attitude as a mild predictor of computer attitude. A study of over 2,000 students in Israel found no link between science achievement and the amount of time spent using a computer (Tamir, 1987). An obvious method for students to increase their computer ability is through enrolling in computer courses. According to research by Campbell and Williams (1990) the most effective computer course for high school students is enjoyable for the students, yet is perceived as useful and has an environment that alleviates the fear of failure.

Wang and Newlin (2002) examined 122 college junior and seniors who enrolled in online sections of a course to determine if their reason for enrolling in a web-based course affected their achievement. Not surprisingly, they found that the students who wanted to enroll in the online sections, as opposed to those who enrolled because it was the only available section, received higher grades in the class. They also found technology and subject content, self-efficacy, and how actively a student used the course website also correlated positively with success in the course. A different study, with a very small population (n=18), found students who indicated they believed class discussion was not helpful were more engaged, as measured by use, on the course website (Moan & Dereshiwsky, 2002). They also suggested learning styles did not correlate with engagement on the website.

Because students at every level often struggle with physics, many studies have been completed to determine what factors contribute to success in a physics course. An investigation of over 400 students in Nigeria correlated variables with physics success (Okpala & Onocha, 1988). The top four correlates were, in descending order, math ability, attitude toward physics, word knowledge, and study habits. Griffith (1985) examined the effects of formal reasoning and math skills on physics achievement in introductory university physics courses. Formal reasoning, math ability, and effort all were strongly correlated with achievement.

Finding variables that consistently account for the variance of success within a population is a difficult task. For example, a study predicting success in an Australian physics course yielded a different list of significant variables from one year to the next (O'Halloran & Russell, 1980). One variable that is consistent in predicting success in physics, however, is math ability (Champagne & Klopfer, 1982; Griffith, 1985; Hudson & McIntire, 1977; Hudson & Rottmann, 1981; O'Halloran & Russell, 1980; Okpala & Onocha, 1988; Wollman & Lawrenz, 1984). In a study of over 900 students who completed a college physics course, math ability by itself was able to account for almost 42% of the variance in the final grade. Interestingly, while math ability correlates very well for success in physics, it does not correlate well with students who drop out of physics courses. Wollman and Lawrenz (1984) and Hudson and Rottmann (1981) both found that math ability was not a significant correlate with the likelihood of a student to drop out of a physics course. This suggests that while math ability is important for a student to succeed in physics, many of the students who are failing to complete physics courses are doing so for reasons other than not having sufficient mathematical backgrounds.

Champagne and Klopfer (1982) correlated variables not to physics achievement in general, but instead to achievement only in the section of physics known as mechanics. They used a variety of assessment instruments to define three variables--math aptitude, science experience, and degree of Newtonian physics understanding. An especially discouraging component of this study was that science experience and Newtonian physics understanding were included as separate variables because:</p> <pre> The finding that the Years of High School Physics variate is not significantly related to the Newtonian physics variables may be taken as an indication that students' exposure to high school physics courses has little effect on their acquisition of the particular knowledge, understanding, and skills that are the components of this variable. (p. 307) </pre> <p>The conclusion of this study, not surprisingly, based on the above statement was that math ability and Newtonian physics were the two variables that correlated significantly with achievement in physics.

Research Intent

This study was unique because not only was it trying to find variables that would correlate with and predict success in a high school physics unit, but it also looked specifically at variables related to the method of delivery of the class. An earlier study by the author indicated that in an university chemistry class that used the World Wide Web (WWW) for homework quizzes and class exams, the students' computer use and comfort were significant predictors of their success (Slykhuis & Banks, 2004). In the present study, the curriculum in the high school physics units was entirely computer-based and partially web-based; therefore, the analysis included several computer related variables. The investigation determined what variables were significant predictors of success in a high school physics unit that utilized computer-assisted instruction. More specifically, this study examined if any computer-related variables were significant, particularly for the subset of the group that completed all the activities online, or were the best predictor variables the traditional variables described in the literature, such as math ability.

Population

This study was completed with 150 students at five different high schools in North Carolina. Ninety-five students at three schools received their instruction in a traditional classroom microcomputer-based laboratory (MBL) setting. These students worked in groups with laptop computers and motion detectors. The teachers in these classrooms actively helped the students to develop their physics understanding. Fifty-five students at two schools received the entirety of their instruction online. These students also worked in groups with computers. The teachers in these classes agreed not to offer instruction on the physics concepts. The students were to develop their understanding from the activities, their peers, and the additional resources found on the course website. Cumulatively, the students ranged in age from 15-18 with 88% of them either 16 or 17. All but one student was either a junior or senior in high school. The students were 56% male, and 53% Caucasian and 35% African-American. One hundred forty-one (96%) of the students had a computer at home, and 136 of those computers were online. A majority of the students, 60%, indicated they used the Internet daily, and only 4% reported using the Internet about once a month. Figure 1 shows the distribution of the students' responses to their computer comfort on a 1-10 (10 is the highest) scale.

[FIGURE 1 OMITTED]

Only 38% of the students had completed physical science, while 89% had completed chemistry. One hundred twenty-one students reported being currently enrolled in a math class, while 145 reported completing a math class last year. Figures 2 and 3 show the distribution of those math courses.

[FIGURE 2 OMITTED]

[FIGURE 3 OMITTED]

Method

The 150 students described above all participated in a 2 to 4-week unit on kinematics. All of the students completed the first six investigations of the Tools for Scientific Thinking-Motion curriculum by Sokoloff and Thornton (1998). This curriculum uses MBLs to teach physics concepts. Prior to instruction, all of the students completed a survey asking for a variety of demographic data as well as information about their computer use and comfort. Students also completed a pre-test of their kinematics physics knowledge, the Test of Understanding Graphs- Kinematics (TUG-K) (Beichner, 1998). This test was developed by Beichner (1994) explicitly to test MBL activities. The KR-20 reliability statistic for the TUG-K was .83, well above the .70 required for a reliable test. The Point-Biserial Coefficient of .74, was well above the .20 required for reliable items. Fifteen science educators established the validity. This same test was used at the conclusion of the unit as a post-test to measure the students' achievement.

A student's score on the TUG-K post-test was used as the dependent variable in this study. The traditional independent variables were age, gender, race, year in school, school attended, prior physical science experience, prior math experience, and score on the TUG-K pre-test. The pre-test was included in the model as a measure of aptitude in this particular topic studied during the unit, kinematics.

Some of these variables were created through a compilation of students' responses given on the survey. If students had previously taken physics, physical science, or chemistry, these were combined into one variable that related their physical science experience. To determine students' math experience, they reported their current math class and the last math class they had completed. The 12 responses given to these questions were placed in one of four categories based on the level of the math class. Each student was then given a category score from 1 through 4 corresponding to the level of the course. The other change was to the category of ethnicity. Since this was an open-ended, self-reported question, a wide range of answers was returned. These were combined into three categories, African-American, Caucasian, and Other.

Variables that were particular to computer use and comfort on the survey included: 1) if there was a computer at the student's primary residence, 2) if that computer was connected to the Internet, 3) if the student had ever taken an online course, 4) how often the student used the Internet, and 5) the student's comfort level with a computer. The variable relating the frequency of Internet use provided the students with five choices--once a month, twice a month, once a week, two or three times a week, or once a day. This was then coded as a 1-5 response. The variable asking them to self-report their computer comfort level was answered on a 1-10 scale. Before the analysis was completed, simple correlations between the numerical independent variables were examined (see Table 1) to determine if any of them were so highly correlated as to be redundant.

Not surprisingly, two sets of variables were highly correlated and logically seemed redundant. A high correlation existed between a student's age (Age) and year in school (Year). Subsequently, age was removed from any further analysis. A high correlation also existed between the variables reporting if a student had a computer at his/her primary residence (Computer) and if that computer was online (Internet). Therefore, the variable Computer was omitted from the analysis and the variable Internet was retained.

Stepwise regression was performed on the data collected from all 150 students who completed the MBL units. Separate regression models were then created for the online group and the classroom group to determine if completing the lab activities entirely online increased the reliance on computer-elated correlates. Lastly, in an effort to repeat prior literature, separate regression models were created, splitting the entire group on the basis of gender to determine if these variables could more accurately explain the variance in achievement of female students.

Results

The stepwise regression to create the model that would explain the most variance was completed in a mixed progression fashion with the significance level for acceptance set at .25. This meant that the variable with the lowest p-value, highest significance, was entered first, and then subsequent variables with p-values less than .25 were added to the model. If at any time, however, one of the variables that had been added to the model no longer remained significant at this level, it would be removed from the model. While these levels of significance are greater than the typical .05 value, they helped generate the most parsimonious model that explains the variance. The model was also restricted to only add whole effects. This meant that for any categorical variable, all or none of the categories had to be added. This method was found to be consistent when crosschecked by a model where variables were added individually in order of their significance to maximize the adjusted [R.sup.2] value.

The first regression model was constructed to account for the variance on the post-test for the entire group (n=150). In this model, three variables were added according to the chosen parameters, pre-test, current math category, and gender. Of these, only two variables were significant at the p < .05 level: pre-test (p < .0001) and current math category (p = .0033) (see Table 2). Gender was the next most significant variable and helps to explain the variance in achievement. Therefore, it was kept in the model as a weak correlate. The full model accounted for 56% of the variance on the post-test.

Next, a similar mixed stepwise regression was applied to the classroom (n=95) and online (n=55) groups separately. This was completed to determine if an increased reliance on computers for instruction, the basis of the online group, is more highly correlated with any computer-related variables than the classroom group.

In the regression model for the online group, more variables were involved, possibly because of the smaller sample size. Table 3 shows all of the variables added to the model. Once again, only pre-test and current math were significant indicators (p < .05) of achievement when initially added to the model. When the model was complete, however, pre-test and gender were the only significant variables. The Current Math category was nearly significant with a p-value of .056. The variable Year, which represents the grade level of a student, was added to the model after the variable Gender, but was removed after the variable Internet Use was added because its level of significance at that point became more than the threshold level of .25. Variables that related the completed math class of the student, Internet access at home, and frequency of

Internet access were all mild contributors to the model. Altogether, this model accounted for nearly 63% of the variance on the post-test by the online group.

For the group that completed the MBLs in a traditional classroom setting, the pre-test score (p < .0001) and current math (p = .021) initially were the only significant (p < .05) variables. At the conclusion of the model, however, Pre-Test, Current Math, and School are significant, with Year just above this level (see Table 4). This model explained nearly 63% of the variance on the post-test by the classroom group.

The identical technique was completed after dividing the group by gender. For the subset of female students (n=66), only pre-test was initially significant (p < .05). The last completed math class, year in school, ethnicity, and computer comfort, were all subsequently added to the model. Table 5 shows the complete model. This model accounted for 54% of the variance on the post-test by female students.

The last multi-directional stepwise regression was completed for the male subset (n = 83) of the entire group. For this group, once again pre-test and current math class were the initial significant indicators of achievement. Unlike other groups, however, those were the only two variables needed to describe the variance in the post-test (see Table 6). These two variables accounted for 56% of the variance on the post-test by the male students.

SUMMARY

For this study, factors were correlated with post-test scores after completing a MBL unit either in the classroom or online setting during a 2 to 4-week period. For the entire group of students, scores on the pre-test and current math classes explained 56% of the variance in the post-test achievement. This pattern was repeated for each of the sub-sets examined. This concurs with the majority of the literature in physics indicating math ability is a primary correlate for physics achievement. It was somewhat surprising that the math variable that was significant was current math instead of completed math. This unit, and the related testing, was completed at the beginning of the school year so students would not have had the opportunity to progress through much curriculum in their current math class. It is not surprising that the pre-test accounted for such a great deal of the variance. A student who shows aptitude in this area of physics, as indicated by a high pre-test score, is also likely to do well on the post-test. Gender was a non-significant (p > .05) correlate with achievement, although male students did score higher than female students for the entire group.

The second focus of this project was to determine if the two groups, with different degrees of reliance on the computer for instruction, had different correlations with achievement specifically regarding computer related variables. Once again, pre-test and math ability were the primary indicators of achievement for both groups. Two computer-related variables were included in the analysis of the online group. Both Internet and Internet Use were non-significantly (p > .05) correlated with achievement on the post-test. Despite the fact that the online MBL unit was completed entirely from a computer and near the beginning of the semester, a student's aptitude in the subject material (kinematics) but not the medium (the computer) appears much more important in determining that student's success than the computer-related variables.

The last focus of this study was to determine if prior research could be repeated and females could be more succinctly correlated with success in physics. This does not seem to be the case in this study. While the model for female students accounted for 55% of the variance in post-test score in the final analysis, five non-significant variables were necessary for this conclusion. The only variable that was initially successful for the female group was the pre-test, which accounted for 44% of the variance. In contrast, 56% of the variance for achievement of the post-test for the male students was accounted for with only two variables, both of which were significant--pretest and current math level. This suggests that whether or not females can be more succinctly correlated with success needs further study.

Overall, these regression tests confirmed achievement in physics is highly correlated with aptitude in the subject, as given by the pre-test score, and mathematical level. Interestingly, the variable of the students' physical science experience was only weakly correlated for female students and did not appear in any other models. For all physics students, regardless of the format of the class, this implies that mathematical preparation continues to be one of the most important characteristics in determining success in a physics class.

References

Adair, R. K. (1991). Using quantitative measures to predict persistence in the natural sciences. College and University, 67(1), 73-79.

Beichner, R. J. (1994). Testing student interpretation of kinematics graphs. American Journal of Physics, 62(8), 750-762.

Beichner, R. J. (1998). Kinematics graph interpretation project. Retrieved February 26, 2004, from www.physics.ncsu.edu:8380/physics_ed/TUGK.html

Campbell, N. J., & Williams, J. E. (1990). Relation of computer attitudes and computer attributions to enrollment in high school computer courses and self-perceived computer proficiency. Journal of Research on Computing in Education, 22(3), 276-289.

Champagne, A. B., & Klopfer, L. E. (1982). A causal model of students' achievement in a college physics course. Journal of Research in Science Teaching, 19(4), 299-309.

Collis, B. (1985). Sex differences in secondary school students' attitudes toward computers. The Computing Teacher, 12(7), 33-36.

DeBoer, G. E. (1985). Characteristics of male and female students who experienced success of failure in their first college science course. Journal of Research in Science Teaching, 22(2), 153-162.

Edge, O. P., & Friedberg, S. H. (1984). Factors affecting achievement in the first course in calculus. The Journal of Experimental Education, 52(3), 136-140.

Griffith, W. T. (1985). Factors affecting performance in introductory physics courses. American Journal of Physics, 53(9), 839-842.

Harackiewicz, J. M., Barron, K. E., Tauer, J. M., & Elliot, A. J. (2002). Predicting success in college: A longitudinal study of achievement goals and ability measures as predictors of interest and performance from freshman year through graduation. Journal of Educational Psychology, 94(3), 562-575.

Hudson, H. T., & McIntire, W. R. (1977). Correlation between mathematical skills and success in physics. American Journal of Physics, 45(5), 470-471.

Hudson, H. T., & Rottmann, R. M. (1981). Correlation between performance in physics and prior mathematics knowledge. Journal of Research in Science Teaching, 18(4), 291-294.

Ignatz, M. (1982). Sex differences in predictive ability of tests of structure-of-intellect factors relative to a criterion examination of high school physics achievement. Educational and Psychological Measurement, 42(1), 353-360.

Levin, J., & Wyckoff, J. (1988). Effective advising: Identifying students most likely to persist and succeed in engineering. Engineering Education, 78, 178-182.

Margrain, S. A. (1978). Student characteristics and academic performance in higher education: A review. Research in Higher Education, 8(2), 111-123.

McCammon, S., Golden, J., & Weunsch, K. L. (1988). Predicting course performance in freshman and sophomore physics courses: Women are more predictable than men. Journal of Research in Science Teaching, 25(6), 501-510.

Moan, E. R., & Dereshiwsky, M. I. (2002). Identifying Factors that predict student engagement in web-based coursework. USDLA Journal, 16(1).

Munday, L. A. (1970). Factors influencing the predictability of college grades. American Educational Research Journal, 7(1), 99-107.

Noble, J. P., & Sawyer, R. L. (1989). Predicting grades in college freshman english and mathematics courses. Journal of College Student Development, 30.

O'Halloran, P. M., & Russell, G. J. (1980). Assessment and prediction of success in first year physics at an Australian university. Higher Education, 9, 529-547.

Okpala, P., & Onocha, C. (1988). Student factors as correlates of achievement in physics. Physics Education, 23(6), 361-364.

Slykhuis, D. A., & Banks, A. J. (2004). Online homework in a university chemistry class Predictors of success and results of modifications to ensure best educational practices. Paper presented at the National Association of Research in Science Education, Vancouver, BC.

Sokoloff, D. R., & Thornton, R. K. (1998). Tools for scientific thinking: Motion and force. Beaverton, OR: Vernier Software & Technology.

Tamir, P. (1987). Some factors which affect science achievement of high school seniors in Israel. Research in Science and Technological Education, 5(1), 69-92.

Wang, A. Y., & Newlin, M. H. (2002). Predictors of web-student performance: The role of self-efficacy and reasons for taking an on-line class. Computers in Human Behavior, 18(2), 151-163.

Wollman, W., & Lawrenz, F. (1984). Identifying potential "dropouts" from college physics classes. Journal of Research in Science Teaching, 21(4), 385-390.

DAVID SLYKHUIS

[e-mail]

James Madison University 

USA

JOHN PARK

[e-mail]

North Carolina State University 

USA

Table 1 Correlation of Independent Variables

                      Year  Comp.  Internet  Online  Internet Use

Age                   .75   -.14   -.21       .05    -.32
Year                        -.06   -.10      -.05    -.31
Computer                            .65      -.10     .20
Internet                                     -.02     .38
Online                                                .16
Internet Use
Computer Comfort
Phy. Sci. Experience

                      Comp. Comfort  Phy. Sci. Exper.  Math  Pre-test

Age                   -.12            .16              -.21  -.27
Year                  -.19            .06              -.18  -.27
Computer               .14            .02               .04   .13
Internet               .17            .04               .04   .21
Online                 .11           -.03              -.07  -.11
Internet Use           .51            .05              -.16   .24
Computer Comfort                      .09              -.00   .13
Phy. Sci. Experience                                    .19   .08

Note: all values rounded to two decimal places

Table 2 Correlates With Post-test for all Students

Variable      Initial p-value  Final p-value  [R.sup.2]

Pre-Test      < .0001          <.0001         .497
Current Math    .003            .005          .555
Gender          .206            .206          .560

Note: The [R.sup.2] values are cumulative as that variable is added to
the model. The initial p-value is the value at the time it was entered.
The final p-value is the p-value after all the variables have been
added.

Table 3 Correlates With Post-test for the Online Group

Variable        Initial p-value  Final p-value  [R.sup.2]

Pre-Test        <.001            .003           .330
Current Math     .016            .056           .506
Gender           .063            .007           .549
Year             .202            Removed        .569
Completed Math   .213            .105           .621
Internet Use     .214            .103           .639
Internet         .189            .189           .647

Note: The [R.sup.2] values are cumulative as that variable is added to
the model. The initial p-value is the value at the time it was entered.
The final p-value is the p-value after all the variables have been
added.

Table 4 Correlates With Post-test for the Classroom Group

Variable      Initial p-value  Final p-value  [R.sup.2]

Pre-Test      < .0001          < .0001        .529
Current Math    .021             .020         .589
School          .127             .020         .609
Year            .055             .055         .627

Note: The [R.sup.2] values are cumulative as that variable is added to
the model. The initial p-value is the value at the time it was entered.
The final p-value is the p-value after all the variables have been
added.

Table 5 Correlates With Post-test for Female Students

Variable          Initial p-value  Final p-value  [R.sup.2]

Pre-Test          < .0001          < .0001        .428
Completed Math      .134             .058         .468
Year                .194             .093         .484
Ethnicity           .146             .084         .520
Computer Comfort    .108             .108         .544

Note: The [R.sup.2] values are cumulative as that variable is added to
the model. The initial p-value is the value at the time it was entered.
The final p-value is the p-value after all the variables have been
added.

Table 6 Correlates With Post-test for Male Students

Variable      p-value initial  p-value final  [R.sup.2]

Pre-Test      < .0001          < .0001        .487
Current Math    .030             .030         .560

Note: The [R.sup.2] values are cumulative as that variable is added to
the model. The initial p-value is the value at the time it was entered.
The final p-value is the p-value after all the variables have been
added.