Morris, John D.

This study explored the predictive accuracy of three, two-group linear discriminant function equations. The grouping or criterion variables employed were scores from the three sections (math, reading, and writing) of the Computerized Placement Test (CPT). The predictor variables used were gathered from the academic history for each case. These variables ranged from the GPA in grade eight, through the normal-curve-equivalent test scores in both math and reading from the Grade Ten Assessment Test (GTAT), and included several actual GPAs achieved in science and math courses at all three levels of difficulty in high school (Appendix A). The data set was built from 16 variables derived from the academic history of 432 graduates of the Broward County public schools in 1994. Each of these cases was found to have had a pass/fail score on the Florida Entry-Level-Placement Test (CPT) at Broward Community College. All three discriminant function equations were examined to determine if the models were statistically significant and reliable, or consistently repeatable, using both a leave-one-out analysis, and a hold-out estimate analysis. The degree to which the models performed better than proportional chance was also investigated. The results of the three, two group discriminant equations built from the predictor variables and the grouping variables have implications for high school counselors. These data indicated a better prediction for students who do not pass (90.4%) the CPT than for those who do pass (48.8%) the entry-level placement test at Broward Community College. The external analyses (leave-one-out and hold-out) show some reduction in predicting accurately; but, as demonstrated by the Z class summary, still are predicting at a significant level for the samples with the exception of the CPTRJ hold-out method for the non-selected group. For high school counselors, then, this research was designed to encourage early examination of particular data sets for the purpose of predicting success on the Florida entry-level placement test. Successful course selection can assist students gain access to college.

The purpose of this study was to provide a predictive model for profiling potential high school dropouts during the middle school years using data available to teachers, counselors, and administrators. The subjects for the study were 377 students (175 dropouts and 202 persisters) in grades 9 through 12 that attended the Okeechobee County Schools between the 1992-93 and 1995-96 school years. The null hypothesis stated that a model consisting of a subset of the 23-predictor variables would not discriminate between students who drop out of school and those who persist. Data were collected from archival sources within the school district and analyzed through discriminant analysis using an SPSS program. Cross validation was performed using a FORTRAN program developed by Morris and Meshbane (1995). Numerous models were identified that discriminated between dropouts and persisters; therefore, the null hypothesis was rejected. The model that had the highest hit rates contained the following variables: (a) age in the eighth grade, (b) suspensions in the eighth grade, (c) attendance in the seventh grade, (d) attendance in the eighth grade, (e) membership in a dropout prevention program, (f) free or reduced-price lunch status, (g) limited-English proficiency status, (h) retention status, and (i) athletic status. The variable that appeared most often in the models that had the highest hit rates for the dropout group was retention status. Other variables that appeared in models included the number of Ds and Fs in the seventh grade, the number of Ds and Fs in the eighth grade, and suspensions in the seventh grade. Variables that were not found to be as significant based on F-ratios included (a) gender, (b) membership in an exceptional education program, and (c) certain achievement test scores. A large amount of missing data relative to the achievement test scores may have contributed to the lack of significance. Hispanic students dropped out in disproportion to the representation in the sample. In addition, most of the dropouts in the sample were not in dropout prevention programs, and most of the dropouts that were in a dropout prevention program dropped out despite the intervention. A profile of the high school dropout based on this study includes: (1) being retained more often, (2) being overage, (3) receiving more Ds and Fs, having poorer attendance, and being suspended more often in the seventh and eighth grades, and (4) receiving a free or reduced-price lunch.

This study investigates the relationships between and among math anxiety level, perceptual learning style (audio, visual, tactile/kinesthetic), age, gender, and math performance. The subjects were 501 community college students taking remedial credit Introductory Algebra and college credit Basic College Algebra. A questionnaire measuring math anxiety level, perceptual learning style, and personal demographics was developed and administered to the participants. Math anxiety level was measured by the Brief Math Anxiety Rating Scale (BMARS), a subscale of Suinn's Math Anxiety Rating Scale (MARS). Perceptual learning style was measured by the Learning Style Inventory-Adapted (LSI-A), an adaptation of the CITE Learning Style Inventory. The results showed that math anxiety level was significantly correlated to one or more learning styles for all groups studied. Math Anxiety level was also significantly correlated to gender but did not have significant correlations with age or math performance. For the female subjects, there were significant positive correlations between math anxiety level and two learning styles: tactile/kinesthetic and audio. For males, there was a significant positive correlation between math anxiety level and audio learning style only. While the math anxiety levels of females were significantly higher than those of males, their course grades were as well. They were also significantly older than the males in the study group and had significantly higher preferences for the visual learning style than the males. Multiple regression analyses were performed with the predictor variables of age, gender, learning style; and the criterion variable math anxiety level. The regression models were statistically significant and predicted up to 15% of the variance in math anxiety level. Multiple regression analyses were performed on subgroups of the original groups. These groups were comprised of the participants who received letter grades from A-F. Those receiving incompletes or withdrawals were not included in this part of the study. The predictor variables for the regressions in these models were: age, gender, math anxiety level, and perceptual learning style. The criterion variable was math performance, measured by the final grade in the course. All but one of the regression models were statistically significant, predicting up to 16% of the variance in math performance. The remedial credit model was not significant. Further research is needed with a more comprehensive learning style instrument and possibly a different measure for evaluating math performance that would also include all the incomplete grades and withdrawals.

The purpose of this study was to examine teacher perceptions about pertinent factors affecting the successful teaching of the National Standards for Music. Subjects were music specialists and fourth grade generalists from public elementary schools in consenting school districts from throughout the State of Florida. The two groups were administered a survey to glean opinions of the feasibility of implementing each of the nine National Standards for Music by rating seven repeated professional and resource items (contact time, resources, assistance, ability, training, interest, responsibility, and level of assistance). The summed ratings of the dependent variables (consisting of the seven professional and resources items) for each standard were subjected to a two way analysis of variance with repeated measures to determine that significant differences exist with respect to the independent variable of teacher role, with music specialists and general educators as two levels of this variable; and the independent variable of content standard, variable. Statistical significance and interactions were studied within content standards and between music specialists and fourth grade generalists. Results indicate that significant differences exist between music teachers and fourth grade teachers in their perceptions of the feasibility of delivering effective instruction for each standard. Additionally, a significant difference exists between each standard with respect to the overall perceived viability of effective implementation regardless of teacher role. Further, an interaction was identified that indicated differences by both role and standard. More specifically, this research yielded results which indicate that certain standards (History & Culture, Singing, and Analyzing Music) are more feasible for both roles to teach than other standards (Playing Instruments, Improvising, and Composing). It also verified that music specialists are considerably more amenable to the implementation of the nine content standards than the general educators with respect to all seven professional and resource items. Music teachers feel most effective implementing the Evaluating, Listening & Analyzing, and Singing standards while generalists feel most effective implementing the History & Culture, Other Subjects, and Singing standards. Both groups indicated an overall lack of time and resources to effectively teach most standards. Music teachers were less dependent on the assistance of classroom teachers while classroom teachers agreed that they needed the assistance of music teachers to successfully implement most standards.

The cross-validated classification accuracy of predictive discriminant analysis (PDA) and logistic regression (LR) models was compared for the two-group classification problem. Thirty-four real data sets varying in number of cases, number of predictor variables, degree of group separation, relative group size, and equality of group covariance matrices were employed for the comparison. PDA models were built based on assumptions of multivariate normality and equal covariance matrices, and cases were classified using Tatsuoka's (1988, p. 351) minimum chi square rule. LR models were built using the International Mathematical and Statistical Library (IMSL) subroutine Categorical Generalized Linear Model (CTGLM), available with the 32-bit Microsoft Fortran v4.0 Powerstation. CTGLM uses a nonlinear approximation technique (Newton-Raphson) to determine maximum likelihood estimates of model parameters. The group with the higher log-likelihood probability was used as the LR prediction. Cross-validated hit-rate accuracy of PDA and LR models was estimated using the leave-one-out procedure. McNemar's (1947) statistic for correlated proportions was used in the statistical comparisons of PDA and LR hit rate estimates for separate-group and total-sample proportions (z = 2.58, a =.01). Total-sample and separate-group cross-validated classification accuracy obtained by PDA was not significantly different from that obtained by LR in any of the 31 data sets for which maximum likelihood estimates of LR model parameters could be calculated. This was true regardless of assumptions made about population sizes (i.e., equal or unequal). Neither theoretical nor data-based considerations were helpful in predicting these results. Although it does not appear from these data to make a difference which classification model is used, use of the method described in this study for comparing PDA and LR models will enable researchers to select the optimal classification model for a specific data set, regardless of data conditions.

This study investigated the degree of agreement parents and teachers in public and private schools had with full school choice in the form of tuition vouchers. The school choice attitude assessment instrument was developed for this purpose and determined to be reliable. The surveys were administered to 416 volunteer respondents. Significant differences were found in the voucher subscale score obtained from parents and teachers. The parents were more favorable to the concept of school choice F(1, 412) = 10.80, p < .00625. Private school parents and teachers were more positive about school choice than public school parents and teachers, F(1, 412) = 46.63, p < .00625. There was also an interaction between role (parent or teacher) and institution (public or private). The interaction was disordinal; private school teachers had a higher mean than private school parents while public school teachers had a lower mean than public school parents, F(1, 412) = 8.45, p < .00625. The demographic variables of education, age, race, and family income were not significantly related to the voucher subsale score. The demographic variables of education, age, race, and family income were not significantly related to the voucher subsale score. Republicans were found to be more favorable to the concept of school choice than Democrats, F(1, 3) = 3.84, p < .00625. Teacher union members were less favorable toward the concept of full school choice than non-union respondents, F(1, 1) = 67.57, p < .00625. The uniform (with additions) voucher was the most popular, chosen by 38.8% of the respondents. Consequently, full school choice programs in the public schools will probably receive more support from parents than teachers. In the future, full school choice programs will probably originate from outside the public school system.

A two-group predictive discriminant analysis was used to develop a model to predict youth gang activity among students in an urban school district in south central United States. The predictors were selected from age, race, gender, grade repeats, reading comprehension achievement test scores, and changes in school and school district (uprootedness). Data were obtained from elementary school records of 285 grade nine students (151 male and 134 female) in two junior high schools. Ethnic minorities comprised 75.4% of the sample. Gang involvement data were provided by the school district security department. Data were analyzed using SPSS-X and a FORTRAN computer program by Morris and Meshbane (1995). All possible subsets of predictor variables were examined using leave-one-out cross-validation accuracy estimates. Although no models classified both gang and nongang students with greater accuracy than proportional chance expectations, a model consisting of gender, grade repeats, and school and school district changes (uprootedness) yielded significant classification accuracy for gang students, (z = 12.07, p <.05).

This study examined a model to predict success on the Advanced Placement Computer Science (APCS) examination. The sample included all students (N = 423) who participated in the APCS program in the Palm Beach County Public School System from 1985 to 1994. Predictor variables consisted of the number of courses taken in specific content areas at the secondary level, semester grades in the APCS course, grade point average, and gender. Multiple regression analysis indicated the significance of these variables in predicting the score on the APCS examination (F (12,280) = 5.848, p $<$.001). Further discriminant analysis identified the most accurate subset of predictors. All students were divided into two groups based on their scores on the APCS examination (pass/fail). The variables that occurred most frequently in the best subsets included the number of semesters taken in advanced mathematics; overall high school grade point average; gender; the grades achieved in both first and second semester in Advanced Placement Computer Science; and the semesters in computers. A model based on these six predictors had the highest (p $<$.01) predictive accuracy of all models studied (67.6% hit rate). Additional study of other independent variables that contribute to success on the APCS examination is needed.

This research study investigated the relationship between the public's knowledge about and attitudes toward year-round education. The variables investigated included the participants' gender, race, age, number of school-age children, marital status, family income, educational level attained, and sources of information about year-round education. The subjects of this study were 418 adults of Southeastern Florida. Palm Beach County is one of the fastest growing school districts in the country with the addition of approximately 5,000 students each year. There is a need to restructure the educational system with year-round education as one of the possibilities. This study examines the effect of year-round education information on community attitudes. The findings showed that demographic variables appeared to have no significant relationship to attitudes toward year-round education and that the treatment had an ineffective influence on participants' attitudes.

This study investigated the effect of using graphic organizers on ninth grade students' achievement in social studies. Participants were 427 ninth grade students; 316 were regularly tracked with Stanford Achievement Subtest Reading Comprehension percentiles of 35% to 85%. One hundred and eleven students were honors and gifted tracked with Stanford Achievement Subtest Reading Comprehension percentiles of 86% to 90% for honors tracked students and 91% to 99% for gifted tracked students. The regular tracked students formed the sample for experiment one which used a 2 x 6 ANOVA to determine relationships between six graphic organizers and/or elaborations and their effect upon student achievement in social studies, grade nine. The honors and gifted tracked students formed the sample for experiment 2 which used a 2 x 2 x 2 ANOVA to determine relationships between two graphic organizers and/or elaborations and their effect upon student achievement in social studies, grade nine. For both experiments a mixed design was used, with strategy as a between-subjects factor and time as a within subjects factor. In addition, academic track (gifted or honors) was a between-subjets factor in experiment 2. Posttest 1 was an immediate recall test. Posttest 2 was a delayed recall test, which was given three weeks after posttest 1. The results indicated that the graphic organizers and/or elaborations SQ3R (experiment 1) and the Structured Conceptual Knowledge Development, with graphic organizers strategy and elaborations (experiment 1 and experiment 2) were the two most effective learning activities used by students (p $<$.0005). The best strategy for retention of information in the social studies ninth grade content area was the Structured Conceptual Knowledge Development, with graphic organizers strategy and elaborations. It was concluded that specific graphic organizers positively affected achievement of ninth grade social studies students. This study also found that the Conventional Method of Teaching: introducing vocabulary, reading chapter questions, reading the chapter and answering the questions in word or sentence form was the least effective way of improving the achievement of ninth grade social studies students.