Research attitudes of African-American graduate students.

Abstract

This two-stage study examined attitudes toward research of African-American graduate students. Stage 1 involved the development and score validation of the Attitudes Toward Research Design Survey (ATRDS). Stage 2 involved comparing ATRDS subscale scores between African-American students enrolled at an institution located in an urban setting and those enrolled in a rural setting. Comparisons were made with regard to research self-efficacy, perceived professional utility of research, and learning preferences. Stage 2 revealed that African-American graduate students enrolled at the urban institution reported moderately higher levels of research self-efficacy than did their rural counterparts.

**********

In recent years, studies conducted on graduate students enrolled in research methodology/statistics classes have been the subject of doctoral dissertations (e.g., Faghihi, 1998), journal articles, and other publication outlets (e.g., Phillips & Russell, 1994). Research in this area has found that many students are extremely frightened about taking such courses (Wilson & Onwuegbuzie, in press). These students deem research courses to be the most difficult in their programs of study, often revealing that they would not have enrolled in these classes if they had not been required to do so (Wilson & Onwuegbuzie, in press). Moreover, research indicates that many students struggle in research methodology/statistics classes, culminating in underachievement and negative attitudes toward research (Onwuegbuzie, 1997), although these attitudes can be improved by same-race and same-sex mentoring (Hargrove & Frierson, 1994). However, with a very few exceptions (e.g., Hargrove & Frierson, 1994), studies in this area primarily have involved Caucasian-American students. In particular, little is known about the experiences of African-American students, despite the fact that the latter tend to attain lower levels of achievement in research courses than do their Caucasian-American counterparts (Onwuegbuzie, 1999).

In studying African-American students enrolled in research methodology/statistics courses, variables that have a cultural context appear to offer promise. These constructs include learning preferences, research self-efficacy, and perceived utility of research methods. To date, these variables have not been considered simultaneously. Yet, studying these variables within the same framework has the potential to broaden our picture of the characteristics and experiences of African-American graduate students enrolled in research courses. Thus, the purpose of the present study was to compare African-American graduate students enrolled at an institution located in an urban setting and those enrolled in a rural setting with respect to research self-efficacy, perceived professional utility of research, and learning preferences.

Method

Participants The sample comprised 79 African-American graduate students attending either at an institution located in an urban setting (n = 45) or in a rural setting (n = 34). Participants from both settings were enrolled in a quantitative-based research methodology course. The majority of students was female (82.28%). Ages of the sample members ranged from 22 to 62 years (M = 29.41, SD = 7.33).

Instruments and Procedures All participants were administered the Attitudes Toward Research Design (ATRD) survey. This instrument, developed specifically for the present investigation, contains four components. The first component elicits demographic information (e.g., gender). The other three components contain sixty-one 5-point Likert-format items (i.e., "1" = strongly agree; "5" = strongly disagree). These components measure research self-efficacy, perceived professional utility of research, and learning preferences. One component contains 25 items that measure research self-efficacy (alpha = .85; 95% confidence interval [CI] = .80, .89). Another component consists of 9 items measuring perceived professional utility of research (alpha = .67; 95% CI = .55, .77). The remaining component contains 27 items which measured aspects of learning preferences (alpha = .53; 95% CI = .37, .67).

Results

Inter-item correlation matrices were used to conduct three principal component analyses with orthogonal rotations. Specifically, the principal component analyses were conducted on (a) the 25 research self-efficacy items, (b) the 9 perceived professional utility of research items, and (c) the 27 learning preference items. The results of each of these exploratory factor analyses are described below.

Factor Analysis 1: Research Self-Efficacy The initial factor analysis of the 25 research self-efficacy items yielded 9 factors using the eigenvalue-greater-than-one role, also known as K1 (cf. Zwick & Velicer, 1986). The scree plot (Zwick & Velicer, 1986) suggested a four-factor solution. Because the K1 rule has been found typically to overestimate the number of factors (Zwick & Velicer, 1986), the four-factor solution was used. An exploratory factor analysis was then conducted that extracted four factors with orthogonal rotation. The four-factor solution explained 51.63% of the total variance in the correlation matrix. Using a cut-off point of 1.501 (Hair, Anderson, Tatham, & Black, 1995), five items failed to load on any factor. Therefore, these five items were removed, reducing the total number of research self-efficacy items to 20.

After removing these five items, a principal components analysis was undertaken. On this occasion, the scree plot suggested a two-factor solution. Therefore, an orthogonal rotation was conducted. The two-factor solution explained 47.03% of the total variance in the correlation matrix. Using the [absolute value of .50] criterion, two items failed to load on any factor. Therefore, these two items were removed, reducing the total number of research self-efficacy items to 18. After removing these two items, another principal components analysis was undertaken. Again, the scree plot suggested a two-factor solution. Therefore, an orthogonal rotation was conducted. Using the [absolute value of .50] criterion, all items loaded on only one factor. A principal components analysis with an oblique (i.e., promax) rotation also was conducted on the 18-item correlation matrix. This led to no changes in the structure. An examination of the trace (i.e., the proportion of variance explained, or eigenvalue, after rotation) revealed that these two factors explained 50.04% of the total variance. This percentage was between 45% (Hanson, Capraro, & Capraro, in press) and 52% (Hanson & Roberts, in press), which represents the average proportion of variation explained by extracted factors in educational and psychological research studies, respectively. The score reliability pertaining to the 18-item research self-efficacy scale was .87 (95% CI = .82, .91). Factor 1 (alpha = .80; 95% CI = .73, .86), containing 8 items that explained 34.54% of the variance, was labeled Perceived Research Understanding and Competence. Factor 2 (alpha = .90; 95% CI = .86, .93), containing 10 items that explained an additional 15.50% of the variance, was labeled Perceived Research Preparedness and Organizational Skills. Table 1 presents the factor pattern/structure coefficients for the obtained solution for the 18 research self-efficacy items. See issue's website <http://rapidintellect.com/AEQweb/fal2003.htm>

These two sets of factor scores were then used to compare the two sample groups. After applying the Bonferroni adjustment, the t-tests revealed that African-American graduate students attending the urban institution (M = 29.47, SD = 4.89) reported statistically significantly (t = 3.30, p < .05) higher levels of perceived research understanding and competence than did those attending the rural institution (M = 25.50, SD = 5.79). The associated effect size of 0.75 was large (Cohen, 1988). However, no statistically significant difference (t = 0.16, p > .05) emerged between urban students (M = 42.80, SD = 7.92) and rural students (M = 42.56, SD = 4.00). Conversely, using the total scores as an overall measure of research self-efficacy, the urban students (M = 72.27, SD = 10.84) reported statistically significantly (t = 2.02, p < .05) higher levels of perceived research understanding and competence than rural students (M = 68.06, SD = 7.63), with a moderate corresponding effect size (0.44).

Factor Analysis 2: Perceived Professional Utility of Research The initial factor analysis of the nine perceived professional utility items yielded two factors using the eigenvalue-greater-than-one rule, as did the scree plot. Thus, the two-factor solution was used. An exploratory factor analysis was then conducted that extracted two factors with orthogonal rotation. The two-factor solution explained 53.92% of the total variance in the correlation matrix. Using a cut-off point of [absolute value of 50], one item failed to load on any factor. Therefore, this item was removed, reducing the total number of perceived professional utility of research items to 8. However, removal of this item, reduced the number of items that loaded on the second factor to two, which was too small to yield reliable scores. Therefore, the two items from the second factor were removed on the grounds that they did not represent a meaningful and trustworthy solution.

After removing these three items, a principal components analysis was undertaken. On this occasion, both the K1 and the scree plot suggested a one-factor solution. The one-factor solution explained 34.23% of the total variance in the correlation matrix. Using the [absolute value of .50] criterion, one item failed to load on this factor. Therefore, this item was removed, reducing the total number of perceived professional utility of research items to 5.

After removing this item, another principal components analysis was undertaken. Again, both the KI and scree plot suggested a one-factor solution. Using the [absolute value of .50] criterion, all five items loaded on this factor. An examination of the trace revealed that this factor explained 53.41% of the total variance. This percentage was between the normal range of 45% to 52% normally found in exploratory factor analyses (Henson, Capraro, & Capraro, in press; Henson & Roberts, in press). The score reliability pertaining to the 5-item measure of perceived professional utility of research was .78 (95% CI = .38, .97), representing an increase of 16.42% in score reliability from the original 9-item scale. Table 2 presents the factor pattern/structure coefficients for the obtained solution for the 5 perceived professional utility of research items. This factor score was then used to compare the two African-American groups. The independent t-test revealed no statistically significant difference (t = 1.14, p > .05) between urban students (M = 21.60, SD = 4.00) and rural students (M = 20.79, SD =2.19). See issue's website <http://rapidintellect.com/AEQweb/fal2003.htm>

Factor Analysis 3." Learning Preferences The initial factor analysis of the 27 learning preference items yielded 9 factors using the eigenvalue-greater-than-one rule. However, the scree plot suggested a four-factor solution. Thus, the latter solution was employed. An exploratory factor analysis was then conducted that extracted four factors with orthogonal rotation. The four-factor solution explained 41.73% of the total variance in the correlation matrix. However, using a cut-off point of [absolute value of .50], 18 items failed to load on any factor. This explained, at least in part, why the score reliability for this scale, cited above, was low. Therefore, these 18 items were removed, reducing the total number of learning preference items to 9.

After removing these 18 items, a principal components analysis was undertaken. On this occasion, the scree plot suggested a two-factor solution. Therefore, an orthogonal rotation was conducted. The two-factor solution explained 55.92% of the total variance in the correlation matrix. Using the [absolute value of .50] criterion, one item failed to load on any factor. Therefore, this item was removed, reducing the total number of learning preference items to 7.

After removing this item, another principal components analysis was undertaken. On this occasion, the scree plot suggested a one-factor solution. Using the 1.501 criterion, all items loaded on this factor. An examination of the trace revealed that this factor explained 37.41% of the total variance. This percentage was smaller than the 45% to 52% proportion of variance explained that is typically found (Henson et al., in press; Henson & Roberts, in press). The score reliability pertaining to the 8-item learning preference scale was .76 (95% CI = .67, .83), representing an Table 3 presents the factor pattern/structure coefficients for the obtained solution for the 8 learning preference items. See website <http://rapidintellect.com/AEQweb/fal2003.htm>

The learning preference factor scores were then used to compare the two African-American samples. The independent t-test revealed no statistically significant difference (t = 0.25, p > .05) in perceived professional utility between urban students (M = 28.22, SD = 4.90) and rural students (M = 27.97, SD = 3.56). Thus, the urban and rural African-American students were similar with respect to learning preferences.

Discussion

The purpose of the present study was twofold. The first purpose involved the development and score validation of the Attitudes toward Research Design Survey (ATRDS), using exploratory factor analysis. In particular, the psychometric properties of the ATRDS scale were assessed (i.e., structural validity, score reliability). The second purpose was to compare subscale scores emerging from the ATRDS between African-American students enrolled at an institution located in an urban setting and those enrolled in a rural setting.

The series of exploratory factor analyses indicates that the ATRDS offers promise as a tool for studying the attitudes and characteristics of African-American students. Two of the three scales contained in the ATRDS, namely, the Research Self-Efficacy Scale and Perceived the Professional Utility of Research Scale, appeared to possess adequate initial psychometric properties. The Research Self-Efficacy Scale yielded two subscales: (a) Perceived Research Understanding and Competence and (b) Perceived Research Preparedness and Organizational Skills. Both subscales yielded score reliability coefficients that were in the .80s and .90s, which are high for measures of affect (Nunally & Bernstein, 1994).

Although the Perceived Research Preparedness and Organizational Skills subscale scores did not discriminate urban and rural African-American graduate students, locality differences were found with respect to Perceived Research Understanding and Competence subscale scores. Specifically, compared to their rural counterparts, the urban students reported moderately higher levels of self-efficacy pertaining to how much they understand research and how competent they deem themselves to be. Also, the urban sample reported moderately higher overall research self-efficacy levels than did their rural peers. Why these rural students reported lower levels of research self-efficacy should be the subject of future investigations. In any case, it is likely that both the rural and urban African-American graduate students in the sample would benefit from mentorship from faculty preceptors, especially those with the same race and sex as the students (Hargrove & Frierson, 1994).

The next step in the development of the Research Self-Efficacy Scale is to administer it to a larger sample of African-American graduate students and then conduct a confirmatory factor analysis to test the two-factor structure of this scale found in the present study. The dimensionality of this scale also should be assessed on other populations such as Caucasian-American and Hispanic graduate students. Future research should consider using the 18-item Research Self-Efficacy Scale component of the ATRDS, or any future revised version, to compare African-American graduate students to Caucasian-American graduate students and to other minority groups. Also, researchers should assess whether research self-efficacy predicts performance in research methodology courses.

The Professional Utility of Research Scale was found to be unidimensional, yielding an adequate score reliability coefficient. Interestingly, scores for these scales did not discriminate the rural and urban study participants. That is, both subgroups were similar with respect to how useful they found research for their professional careers. Again, future confirmatory factor analysis and other validation techniques are needed to assess further the psychometric properties of this scale.

Finally, the Learning Preference Scale clearly needs more work. Although the score reliability coefficient was adequate, the low proportion of variance explained by the eight items included in the final solution (i.e., 37.41%) suggests that more items are needed. Indeed, the Spearman-Brown prophecy (Crocker & Algina, 1986) predicts that doubling the number of items on the Learning Preference Scale from 8 to 16, could increase score reliability from .76 to .86. This would represent a 13.2% increase in score reliability. Although no difference was found in learning preference scores between the urban and rural African-American graduate students, replications are needed to assess the reliability of this finding.

The present study has made a contribution to the literature by studying a graduate population enrolled in research courses who have been found to experience lower levels of achievement in these courses relative to their peers (Onwuegbuzie, 1999). Moreover, an instrument containing several scales (i.e., ATRDS) was developed that appears to be a useful tool. The major finding in this study is that research self-efficacy has a regional context. This suggests that there might be as much or even more within-race variations as there are between-race variations in research methodology courses.

References

Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, N.J.: Lawrence Erlbaum.

Crocker, L., & Algina, J. (1986). Introduction to classical and modern test theory. Orlando, FL: Holt, Rinehart, and Winston.

Faghihi, F. Y. (1998). A study of factors related to dissertation progress among doctoral Candidates: Focus on student research self-efficacy as a result of their research Training and experiences. Dissertation Abstracts International Section A: Humanities & Social Sciences, 59(5-A), 1456. US: University Microfilms International.

Hair, J. F., Anderson, R. E., Tatham, R. L., & Black, W. C. (1995). Multivariate data analysis (4th ed.). Englewood Cliffs, N J: Prentice Hall.

Hargrove, B. K. & Frierson, H. T. (1994). Assessing a summer research program's impact on African American students' attitudes toward research and perceptions of faculty preceptors. In S. Hood and H. T. Frierson (Eds.) Beyond the Dream: Meaningful program evaluation and assessment to achieve equal opportunity at predominantly White universities (pp. 47-60), Greenwich, CT: JAI Press.

Onwuegbuzie, A. J. (1997). Writing a research proposal: The role of library anxiety, statistics anxiety, and composition anxiety. Library and Information Science Research, 19, 5-33.

Onwuegbuzie, A. J. (1999). Underachievement of African-American graduate students in research methodology classes: Possible implications for the supply of school administrators. The Journal of Negro Education, 67, 67-78.

Phillips, J. C., & Russell, R. K. (1994). Research self-efficacy, the research training environment, and research productivity among graduate students in counseling psychology. Counseling Psychologist, 22, 628-641.

Wilson, V. A., & Onwuegbuzie, A. J. (in press). Changes in anxiety among doctoral students. Academic Exchange Quarterly.

Zwick, W. R., & Velicer, W. F. (1986). Factors influencing five rules for determining the number of components to retain. Psychological Bulletin, 99, 432-442.

Dianne Richardson, The University of West Alabama Anthony J. Onwuegbuzie, Howard University, Washington, DC

Richardson is an assistant professor. Onwuegbuzie is an associate professor.