Search

This paper discusses the basics of repeated measures designs. Within-subjects designs are compared to between-subjects designs, discussing the advantages and disadvantages of each. Further discussion compares a univariate one-way analysis of variance (ANOVA) with the between-subjects ANOVA and multivariate repeated measures ANOVA. Limitations of the univariate repeated measures ANOVA and their corrections are explained. This paper also demonstrates that the univariate repeated measures ANOVA is a form of linear regression. The advantages of linear regression over ANOVA are discussed briefly. The discussion concludes with examples of how to compute univariate, multivariate, and linear regression ANOVAs using the Statistical Package for the Social Sciences. Five appendixes contain tables of study results. (Contains 5 tables and 27 references.) (SLD)

Students successful in developmental coursework later do as well or better than students not needing developmental courses. This study examined the performance of 3,873 students new to a college for the fall 1997 semester through the spring 2001 semester. Analysis of variance and logistic regression were used to tease out effects of level of developmental need and age when starting. Consecutive semester attendance and number of courses dropped were significant predictors for degree or certificate completion. Age group interacted with a number of variables examined in the study. (Contains 7 figures, 17 tables, and 19 references.) (Author/SLD)

This simulation study focused on the power of detecting group differences in linear growth trajectory parameters within the framework of structural equation modeling (SEM) and compared this approach with the more traditional repeated measures analysis of variance (ANOVA) approach. Three broad conditions of group differences in linear growth trajectory were considered. SEM latent growth modeling consistently showed higher statistical power for detecting group differences in the linear growth slopes than repeated measures ANOVA. For small group differences in the growth trajectories, large sample size (e.g., N>500) is typically required for adequate statistical power. For medium or large group differences, moderate or small sample size is usually sufficient for adequate power. Some future research directions are discussed. (Contains 2 figures, 8 tables, and 16 references.) (Author/SLD)

The effectiveness of diversity training in eliminating racial stereotypes in the workplace and modifying employees' negative attitudes toward diversity was examined in a study conducted at a private nonprofit college in the San Francisco Bay area. The study sample consisted of 80 nontraditional students from 4 sections of a course titled Managing Cultural Diversity (MCD) and 40 students from 2 marketing classes. The Cross-Cultural Adaptability Inventory (CCAI) was administered to all 120 students as a pretest and posttest. The MCD students received 7 weeks of intensive training on cultural diversity, interracial differences, race relations, ethnicity and attitude, psychographics, prejudice and discrimination, and white privilege. The 40 marketing students received no diversity training during the study. An analysis of the students' performance on the CCAI using descriptive statistics and repeated measure analysis of variance established that there were no significant differences between the pretest and posttest scores of the two groups. Thus, diversity training did not appear to have altered the attitudes or behaviors of students who took the MCD course. Several possible explanations for this finding were offered, including the fact that the students were already fairly sophisticated with respect to the subscales measured by the CCAI. (Contains 27 references.) (MN)

Unfortunately, researchers do not usually have measurement instruments that provide perfectly reliable scores. Therefore, the researcher may want to account for the level of unreliability by appropriately increasing the sample size. For example, the results of a pilot study may indicate that a particular instrument is not as reliable with a given population as it has been with other populations. A series of Monte Carlo analyses were conducted to determine the sample sizes required when measurements are not perfectly reliable. The methods investigated were: (1) Pearson correlation; (2) Spearman rank correlation; and (3) analysis of variance (ANOVA). Using this information, a researcher can use the tables provided to determine an appropriate sample size for their study. Tables are also provided to illustrate the reduction in power from decreased reliability for given sample sizes. The computer program will be made available through the World Wide Web to help researchers determine the actual statistical power they can expect for their studies with less than perfect reliability. (Contains 1 figure, 6 tables, 6 charts, and 27 references.) (Author/SLD)

This study investigated cognitive style as a screening mechanism for placing students in gifted education programs. Middle school (n=156) and high school students (n=94) were included in this study. Slightly fewer than half of the students had been previously identified as academically gifted, while the remainder were from the regular studies program. Approximately half of the students were female. Participants completed the Group Embedded Figures Test (GEFT), an instrument designed to measure cognitive style. A factorial analysis of variance showed no significant interactions between program type and school year on scores from the GEFT. Main effects, however, were significant for gender and program type. Students identified as gifted scored significantly higher than their counterparts in the regular education program, and males in both programs scored significantly higher than their female classmates. The data were entered into a logistic regression procedure to determine the probability of predicting a given student's program type based on his or her gender and GEFT score. The statistical software removed the gender variable and was able, using the GEFT score, to predict program type in 74% of the cases. Some cautions based on study design and the fact that cognitive style may change with age are discussed. (Contains 1 table and 12 references.) (SLD)

This paper reviews issues involved in converting continuous variables to nominal variables to be used in the OVA techniques. The literature dealing with the dangers of dichotomizing continuous variables is reviewed. First, the assumptions invoked by OVA analyses are reviewed in addition to concerns regarding the loss of variance and a reduction in score reliability that result from the conversion of continuous variables to nominal variables. Second, regression is discussed as a more adequate alternative to analysis of variance (ANOVA). Finally, a heuristic data set is presented and used as a demonstration of three alternatives that can occur through this conversion and when ANOVA is used. (Contains 6 figures and 20 references.) (Author/SLD)

Profiles of 104 welfare recipients in Georgia were examined to identify social indicators and cognitive variables that influenced work role participation. Three instruments were administered the Career Thought Inventory, the Career Decision-Making Self-Efficacy Scale-Short Form, and the demographic profile and participation scale of the Salience Inventory to determine how well race, education, dysfunctional career thoughts, and career decision-making self-efficacy predicted work role participation. It was found that, as a group, the welfare recipients in the sample defied stereotypes; they were older, better educated, and more heterogeneous than is typically reported. Among other findings were the following: (1) a majority of participants indicated at least some difficulty with dysfunctional career thoughts; (2) career decision-making self-efficacy was lower than that of a normative sample of college females; (3) black respondents spent more time in work activities than their white counterparts; (4) non-high school completers indicated more decision-making confusion and commitment anxiety, but less career decision-making self-efficacy than high school completers. It was concluded that career self-efficacy was the most important predictor of work role participation. It was suggested that service providers do the following: (1) incorporate coping mechanisms that recipients could employ to minimize obstacles to self-efficacy into employment intervention programs; and (2) employ awareness and sensitivity to the clash between the future-oriented nature of career/job training and day-to-day survival needs confronting welfare recipients.(There are three tables. The bibliography lists 51 references.)(AJ)

Because an instructor's first interaction with students on the first day of class can determine the success of those to follow, it is important to explore what happens in a classroom setting before offering prescriptions on how to best handle situations in that setting. To understand how students' responses to specific attributes related to instructional style are linked to global evaluations of teaching effectiveness, six research questions were formulated. Participants were 800 students at the University of Illinois--125 were freshmen, 434 were sophomores, 140 were juniors, 79 were seniors, 8 were graduate students, and 14 did not indicate their rank. Data were collected in 29 sections of three 100-level courses in three different subjects. To gather reactions to the first day of class, a multi-part survey was developed. The section of the questionnaire that provided primary data for analyses consisted of 22 five-point Likert Scale items. Findings suggest that the most important element of instructional effectiveness on the first day of class is communicative competence. Since the first day of class is a day where students begin to develop an instructional relationship with their teacher, it appears they are looking for communicative competence to reassure them that their instructor has attributes that will enable them to succeed in the course. Demographic and attitude elements such as a student's year in school, an instructor's gender, prior student motivation, and even course type do appear to play some role in how a student evaluates an instructor on the first day. (Contains 25 references and 9 tables.) (NKA)

This monograph attempts to synthesize and interpret the extant research from the last 4 decades on the impact of schooling on students' academic achievement. The central thesis is that educators stand at the dawn of a new era of school reform. The discussion, which is somewhat technical in nature, relies on five indices to describe the relationship between student achievement and various school, teacher, and student-level factors. These are: (1) percent of variance explained; (2) the correlation coefficient; (3) the binomial effect size display; (4) the standardized mean difference effect size; and (5) percentile gain. The first section of the paper includes the introduction and chapters 2 and 3, which review the literature on previous attempts to identify the variables impacting student achievement. The second section, chapters 4, 5, and 6, presents a discussion of the research on school level variables. The final section, chapter 7, considers the implications of the findings for school reform. Findings indicate that schools can influence student achievement profoundly. The conclusions suggest that student achievement can be affected strongly if schools provide teachers with well-articulated curricula. They should optimize their use of instructional time, establish achievement goals for students and monitor those goals, and they must communicate a clear message that high academic achievement is the primary goal of the school. It is important to involve parents, maintain an orderly and cooperative environment, and involve staff in all key decisions. (Contains 38 tables and 152 references.) (SLD)

Step-down analysis is a multivariate technique that examines dependent variables across groups by using a series of univariate "F" tests done in an a priori order. The first "F" test results are the same as a univariate "F" test examining the dependent variable. Each of the following "F" tests then uses the previously used dependent variables as covariates. The null hypothesis tested by this method is that there is no effect across groups when dependent variables are listed in a specific order. Because step-down uses covariates, as does analysis of covariance (ANCOVA) and multivariate analysis of covariance (MANCOVA), and is performed in a series of steps, somewhat like stepwise analysis, step-down is compared across these different methods. The comparisons show that step-down analysis does not have the same problems as ANCOVA, MANCOVA, and stepwise analysis. However, step-down has its own problems that are discussed in more detail. (Author/SLD)

Many people use post hoc tests but do not completely understand why regular t-tests are not used post hoc or exactly what these tests are doing. This paper makes a direct comparison of Tukey and Scheffe post hoc tests with regular t-tests conducted using a new testwise alpha to make clear that controlling experimentwise error through some kind of Bonferroni-type correction is the basic concept underlying post hoc methods. (Contains 11 references.) (Author/SLD)

The analysis of variance (ANOVA) is a frequently used statistical procedure by which the equality of more than two population means can be tested without inflating the Type I error rate (D. Hinkle, W. Wiersma, and S. Jurs, 1998). Fixed-, random-, and mixed-effects ANOVA models are each capable of yielding interesting and useful results when applied in appropriate situations. However, the random- and mixed-effects models offer the added benefit of increasing the generalizability of results. This paper illustrates "rules of thumb" (C. Hicks, 1973) researchers can use to test all three models, explores the factors that should bear on model selection (B. Frederick, 1999), and explains how the Statistical Package for the Social Sciences (SPSS) can be use to evaluate all three models. An appendix contains the SPSS program to analyze fixed- and random-effects models. (Contains 3 tables and 19 references.) (Author/SLD)

This study examined the effect of the gender specific aspects of a child's clothing and a child's sex on the type of interaction that occurred between the child and the teacher. A stratified random sample of second graders, including urban (n=104), suburban (n=77), and rural (n=49) districts, was used, drawing on two classes from each of six school districts. Principals and teachers were told that the purpose of the study was to observe children's clothing practices and their interactions during class. No mention was made concerning the teacher's interaction with the child. Sex and gender aspects of clothing were found to be related to teacher student interaction. Male and female children who were dressed in a unisex manner were the most likely to experience positive interaction while those dressed in the most gender specific manner were the most likely to experience reproof. In keeping with the findings of previous studies, male students rated significantly higher than female students in interaction with their teacher. (Contains 2 tables and 17 references.) (BT)

A study tested a method for developing reading fluency with third-grade Hill Tribe children in a welfare school in Chiang Mai, a city located in northern Thailand. Most of the students were bilingual, speaking their native tongues and Thai, their second language. Only 18.7% of the Hill Tribe population can read Thai, and many students fail to complete school because of difficulties in reading Thai. Consequently, finding improved ways to develop reading fluency with this population is important--one way to increase fluency is through repeated reading. Repeated reading and self-directed behavior were tested with control conditions to determine if these methods would be effective in promoting reading fluency in Thai. Samuels' (1978) unassisted method was used. All the 95 students in the boarding school were tested to select poor readers. Participants were 28 students who met the criteria, and selected for a balance of sex, tribe, and reading skills such as speed, comprehension, and error detection. Students were randomly assigned to one of four treatments. Analysis of variance was computed for each of the variables to ensure that all groups were equal at the experiment's start. Data analysis showed that students in the repeated reading conditions read texts significantly faster than students in the non-repeated reading conditions and were also significantly better at detecting spelling errors in the text than students in the non-repeated reading conditions. Differences were not found between the groups in comprehension. Findings suggest that students acquire automaticity and fluency for the materials practiced during the repeated reading practice. (Contains 4 tables and 24 references.) (NKA)

This obtrusive post-hoc quasi-experimental study investigated Scholastic Assessment Test (SAT) scores of 111 high school students in grades 10 through 12. Fifty-three students were enrolled in at least one Advanced Placement (AP) course at the time of the study. General factorial analysis of variance (ANOVA) tested for significant differences between verbal, quantitative, and composite SAT scores with the independent variables of class enrollment (AP, non-AP), economic level (high EL, low EL), and gender. Economic level depended on student qualification for the National School Lunch Program. Analysis of the data indicated that there were significantly higher verbal (p=0.001), quantitative (p=0.038), and composite (p=0.003) SAT scores for students enrolled in the AP program. In addition, ANOVA of two-way interaction between class enrollment and economic level indicated that high EL students enrolled in the AP program had significantly higher quantitative SAT scores (p=0.019). All other variable interactions produced no significant differences, although females did have higher scores than males. (Contains 21 tables and 13 references.) (Author/SLD)

This study addressed whether "reading-as-the-reader" can improve fifth- and ninth-grade students' abilities to compose descriptive writing consistent with their readers' informational needs. There were 206 participants: 154 writers (78 fifth graders and 76 ninth graders) and 52 ninth-grade readers. The study adapted the referential communication design from Traxler and Gernsbacher (1992 and 1993) to investigate whether young writers can benefit from a perspective-taking task as they compose and revise their descriptions of tangrams (Chinese geometric puzzles) over three separate writing sessions. Three perspective-taking conditions were contrasted: a numerical feedback condition, a "rating other" description condition, and a "reading-as-the-reader" condition. Readers' correct description-to-tangram matches made for each session served as the dependent measure. Two analysis of variance (ANOVA) were conducted; the first contrasting sessions one and two, the second contrasting sessions one and three. The first analysis revealed a main effect of session. The second analysis revealed a significant session-by-task interaction. Post hoc analyses of the interaction (Tukey's HSD) established that only the read-as-the-reader group showed a significant gain between sessions one and three. A discourse analysis of the descriptions revealed that students used an "object-centered" discourse strategy that began with an analogy followed by geometric shape and spatial characterizations. The study offers evidence that "reading-as-the-reader" can help younger writers in considering the multiple functions of writing and the reciprocity between writers, readers, and written text. (Contains 36 references, 4 tables, and 2 figures of data.) (Author/RS)

Analysis of variance (ANOVA) designs typically use what is referred to as crossed design to test for differences between means of groups. In a balanced, or crossed, one-way ANOVA, each student (unit of analysis) would have a score in each of the experimental conditions. In a two-way design, the analysis is considered crossed if each level from one way is contained in each level of the other way. In this design, every person (unit of analysis) has a score in every cell. Many experimental designs in the behavioral sciences do not qualify as a crossed design. Units of measurement are rather "nested" inside other factors. A crossed experimental design would neglect the hierarchical structure of the data and produce incorrect interpretations of results. Neglecting a nested design when one actually exists will make the researcher: (1) wrongly attribute a main effect to an interaction effect when, in fact, no interaction exists; (2) divide by the wrong degrees of freedom when determining the mean square and F-value (and the statistical significance of the F-value); and (3) assume that a main effect has a smaller effect size (eta-squared) because the sum of squares for that effect is being partly attributed to the interaction effect. (SLD)

This paper reviews the basics of the elusive but important concept of the interaction effect. The analysis of variance (ANOVA) interaction effect is defined both conceptually and mathematically. Common errors in interpreting interaction effects are discussed, and appropriate strategies for achieving post hoc understandings of the origin of detected interaction effects are presented. It is important that researchers in the fields of education and psychology learn to report and interpret interaction effects correctly when using an ANOVA design. (Contains 16 references, 4 tables, and 3 figures.) (Author/SLD)

This paper analyzes what happens to the effect size of a given dataset when the variance is removed by categorization for the purpose of applying "OVA" methods (analysis of variance, analysis of covariance). The dataset is from a classic study by Holzinger and Swinefors (1939) in which more than 20 ability test were administered to 301 middle school students to discern which abilities determined overall academic performance. One-way analyses of variance (ANOVA) were run with the dichotomized variable as the independent factor to compare the reduction in effect size in the ANOVAs with the effect size of regression run with the intact continuous independent variable. The effect sizes of the ANOVAs that were run with the dichotomized variable were about half of the regression effect sizes run with the unaltered independent variable. Each of the effect sizes after dichotomization would fit into a lower category, resulting in a serious underestimation of the substantive effect. (Contains 3 tables and 16 references.) (SLD)

The purpose of this study was to determine if there were differences in evaluation of an educational interactive video program between high school students and their parents over a 4-year time span. High school students participating in the interactive video program were surveyed during 5 semesters. Survey instruments were also sent to parents via the student. There were 1,024 student respondents and 481 parent respondents. Constructs for each group were determined using exploratory factor analysis. The mean vectors of similar constructs for each status group were contrasted using multivariate analysis of variance (MANCOVA) with semester and site as additional factors. In addition, similar questions for each status group that would not be included in a construct were also contrasted by MANOVA. Because heterogeneity of covariance matrices was not achieved in the initial sample, a stratified random sample of 150 was selected from each site and status group and the analysis repeated. Results indicate there are no differences in evaluation of the interactive video program between parents and students. These results are discussed. (Contains 4 tables and 19 references.) (Author/SLD)

This study was conducted to determine if there were differences in strengths and weaknesses of an educational interactive video program perceived by high school students and their parents over a 4-year time span. Parent and student responses to open-ended questions concerning strengths and weaknesses of the interactive distance learning program were first categorized and contrasted by status (student or parent). There were 1,505 respondents (1,024 students and 481 parents). There were no differences in perceived strengths. There were, however, differences detected in weakness responses based on status. Loglinear models were then used to examine strength by status by site (home/remote), strength by status by semester, weakness by status by site, and weakness by status by semester. Statistical significance was detected in all models. Results are discussed. (Contains 6 tables, 6 figures, and 18 references.) (Author/SLD)

This report examines how class size and course motivation are related to evaluation results, its purpose being to provide a comparison of rating differences on the Instructor and Course Evaluation System (ICES). ICES offers faculty a computer-based catalog of over 1,000 survey items. The article is based on a study that sought to extend the ICES research by using an unbalanced nested ANOVA to analyze the effects of five factors: course motivation, course level, class size, academic discipline, and individual class differences on (1) student ratings of instructors; and (2) student ratings of overall course quality. The results of the ANOVA showed that course motivation, course level, class size, discipline, and individual class do influence student ratings of both teaching effectiveness and course quality. Different combinations of course motivation, level, and discipline displayed different patterns of student ratings. The significant two-way interactions (in which one of the factors alters the effect of the other) were course motivation and discipline, course level and class size, and course level and discipline. Generally, ratings of elective courses were higher than those of mixed courses, and ratings of mixed courses were higher than those of required courses. Higher level classes had higher ratings than lower level ones. Contains 27 references, 10 figures, and 9 tables. (RJM)

This experimental study investigated internal (psychological characteristics) and external (World Wide Web site features) factors influencing learning and disorientation in Web navigation. The research design was a two-factors ANOVA (ANalysis Of VAriance) with mode of navigation (linear, nonlinear) and distracters (i.e., the presence of "seductive" links) as the two factors. Volunteer participants (n=75) were recruited from Internet listings that advertised the study, drawing from a diverse nationwide population. This paper reports findings of the following dependent measures as they relate to the two-factor research design: friendliness of the Web site, attribution of disorientation, overall level of disorientation, confidence in navigating the Web site passage, interest in Web site passage topic, accuracy at stating main point of the passage, and number of ideas recalled from content of passage. It was found that preference for sensation-seeking behavior and/or spatial-holistic ability did not influence the participants' perceptions of disorientation. Other results indicated that recalling examples from the text was facilitated when the site had no distracters, but an interaction indicated that participants' conception of the main point of the passage was negatively influenced if the site had no distracters when in the nonlinear navigation mode. (Author/MES)

An earlier review by the author reported an aptitude-treatment interaction (ATI) relating instructional feedback and prior knowledge. For low prior knowledge learners, single-try feedback (STF) was slightly superior to multiple-try feedback (MTF), while for high prior knowledge learners, MTF was superior to STF. In the current report, raw score data from existing studies that examined STF and MTF, but that had not considered the possibility of an ATI, are combined and reanalyzed by ANOVA (ANalysis Of VAriance). This study addresses the following questions: is there an interaction between learner prior knowledge and number of feedback tries provided by computer-based training (CBT) lessons; and, if so, how large is this effect? The results of this analysis are presented along with a tentative explanation of the observed effect. Prescriptive application of the ATI and suggestions for future study are provided. (Author/MES)

Aptitude-treatment interaction (ATI) studies have been used with some frequency, yet many researchers do not understand fully what interaction effects are. Because the means for interactions involve fewer persons per mean, power to detect interaction effects is typically smallest for the highest-order interaction in a given design. This phenomenon has been formalized by some methodologists as the Type IV error--the failure to detect statistical significance for the interaction null hypotheses that really should be rejected. This paper reviews the concept of the interaction effect. Small heuristic data sets are used to make the discussion more concrete. The concept of interaction is important because it links the two-way factorial designs with designs involving three or more independent variables. An appendix illustrates the analysis of variance technique. (Contains 4 figures, 5 tables, and 12 references.) (Author/SLD)

A study examined the longstanding theory of the Bavelas group which suggests that the only consistent cause of interpersonal equivocation is avoidance-avoidance conflict (AAC). The study also aimed to uncover a psycholinguistic profile of equivocation, especially in the form of paralinguistic cues such as disfluencies. Subjects, 18 college undergraduates, responded orally to questions from hypothetical interlocutors within scenarios designed to manipulate both the presence/absence of AAC and level of situational formality. Their responses were audiotaped, transcribed, rated for degree of equivocation, and coded for disfluencies. Results of ANOVA showed that AAC did result in equivocation, but also that formality level interacted with AAC in influencing equivocation. Subjects used filled pauses, surprisingly, in the condition within which they equivocated the least, although they produced other disfluencies (combined) within conditions where they equivocated the most. Findings demonstrate that, with regard to the antecedents of equivocation, the Bavelas group was right in claiming that AAC routinely leads to increased equivocation. Future research should aim for a clearer conceptual definition of "formality," and the search for antecedents should extend to individual differences as well. A more reliable method of rating degree of equivocation (as defined by Bavelas) must be used, employing more trained judges. With regard to the psycholinguistic profile of equivocation, the notion that both filled pauses and false starts might provide a substantive index of AAC-based equivocation are not supported. (Contains 2 tables of data and 12 references; an appendix contains "experimental stimuli.") (NKA)

This study examined graduation and attrition patterns of undergraduates who were enrolled in the fall semester but who did not return for the spring semester at a doctoral granting, metropolitan university. A total of 504 of the 1,262 students who enrolled in the fall 1992 semester but who did not return for the spring 1993 semester completed a questionnaire on persistence and attrition. Of the 504 stopouts, 482 were tracked using institutional enrollment files through the fall 1997 semester. As of the fall 1997 semester, 166 of the 482 stopouts were enrolled or had graduated from the institution, while 316 had not returned. It was found that 50 percent of the variance in freshmen persistence and attrition behavior among the stopouts was explained using a regression model containing 13 independent variables. It was also found that 16, 6, and 5 independent variables predicted 70 percent, 32 percent, and 24 percent, respectively, of the variance in persisting and nonreturning behavior among sophomores, juniors, and seniors. Two appendixes provide a copy of the nonreturning students' survey form and survey item correlations. (Contains 20 references.) (MDM)

This paper tested the hypothesis that certain cognitive, affective, and psychomotor variables were significant predictors of college persistence in the first few weeks of the freshman year. Data for the study involved 462 freshmen, with an average age of 18, who matriculated in the fall of 1997 at a public higher education institution; about 93 percent of the participants were male; the ethnic background of the students was 8l.5 percent white, 8.4 percent black, 5.4 percent Asian, and 3.3 percent Hispanic. About 32 percent of the participants were engineering majors, 17 percent were science majors, and 47 percent were majoring in liberal arts. Dropouts were classified into three groups and were then compared with persisters in an analysis of variance study. Logistic regression identified self-efficacy and physical fitness as positive predictors of freshman retention, while judgment and empathy were negatively associated with persistence. Three reasons were identified for freshman attrition: inability to handle stress, mismatch between personal expectations and college reality, and lack of personal commitment to a college education. Intervention strategies to prevent early withdrawal are also discussed. Four tables summarize some of the data. (Contains 22 references.) (CH)

This study represents a continuation of research efforts to further refine the Attitudes & Beliefs on Classroom Control (ABCC) Inventory. Formerly titled the Inventory of Classroom Management Style, the ABCC is an instrument designed to measure teachers' perceptions of their classroom management beliefs and practices. It is based on a continuum originally suggested by C. Wolfgang and C. Glickman (1980, 1986). Objectives were to: (1) investigate the impact of classroom management training on classroom management style; (2) study the relationship between class size and classroom management style; (3) investigate differences between the perceived classroom management style of teachers who had and had not enrolled in a graduate course within the preceding 6 months; and (4) further substantiate the construct validity of the ABCC. Data were collected from 281 certified teachers, who were primarily urban, and female. Most of the teachers were Caucasian (69.9%) and they had an overall average of 14.35 years of teaching experience. Results show significant differences on the Instructional Management subscale of the ABCC regarding classroom management training as well as significant positive correlations between average class enrollment and teachers' scores on the People Management and Behavior Management subscales of the ABCC. A one-way analysis of variance did not yield significant differences between the teachers who had enrolled in graduate courses in the last 6 months and those who did not. Results seem to be in keeping with the construct. (Contains 3 tables and 26 references.) (SLD)

The conventional two-group differential item functioning (DIF) analysis is extended to an analysis of variance-like (ANOVA-like) DIF analysis where multiple factors with multiple groups are compared simultaneously. Moreover, DIF is treated as a parameter to be estimated rather than simply a sign to be detected. This proposed approach allows the investigation of the effects of DIF on items more thoroughly. Results of simulation studies show that the parameters of the proposed models were recovered very well. A real data set with 10 dichotomous items was analyzed. Implications and applications are discussed. (Contains 4 tables, 2 figures, and 10 references.) (Author/SLD)

J. Barnette and J. McLean (1996) proposed a method of controlling Type I error in pairwise multiple comparisons after a significant omnibus F test. This procedure, called Alpha-Max, is based on a sequential cumulative probability accounting procedure in line with Bonferroni inequality. A missing element in the discussion of Alpha-Max was the empirical determination of actual probabilities of Type I errors. This paper compares the Type I error rates of Alpha-Max with other commonly used multiple comparison procedures: (1) Fisher's Least Significant Difference (LSD); (2) Dunn-Bonferroni; (3) Tukey's Honestly Significant Difference (HSD); (4) the Student Newman Keuls (SNK) procedure; and (5) the Scheffe approach. Monte Carlo procedures were used to generate 10,000 replications with varied alpha of 0.05 and 0.01; 3, 4, and 5 groups; and 5 sample sizes. Actual Type I error rates were determined for the greatest difference and for total number of Type I errors. Results indicate that in virtually every situation LSD and Alpha-Max had significantly higher probability of Type I errors than the other four methods. SNK and HSD had higher than nominal alpha probabilities for committing Type I errors, with SNK having a lower level than HSD. Dunn-Bonferroni had a level slightly lower than the nominal level, while the Scheffe had a level much lower than the nominal level. Varying sample size had little practical significance. While Alpha-Max did not provide for acceptable experiment-wise control of Type I error, it may provide an alternative for control of Type I error in the planned, nonorthogonal situation or in situations where assumptions of analysis of variance are violated. (Contains 25 tables and 4 references.) (Author/SLD)

In light of the continuing debate over the relative merits of various ways of assessing student achievement, it seems appropriate to examine longitudinal data which reflect student performance on one measure as a basis for recognizing and utilizing the measure's particular strengths. This study is the follow-up to an earlier study conducted using data from the state of Tennessee for student scale scores in science for the years 1990-1994. The current data set consisted of school system level science scale scores on the CTBS/4 science test, grades 2 through 8, for each of the 138 Tennessee school systems for the years 1992-1996. The normed portion of the science subtest of the CTBS/4 consists of 20 items with four possible answers for each item. Examination of the descriptives revealed an increase in the mean of science scale scores for grades 2-8 each year except 1993, which showed a "negative gain." This population represents the remaining members of the cohort of students (grades 4-8) included in the mean score for 1991, which also had a "negative gain." This finding seems to indicate that teacher effect on student achievement may be both cumulative and residual. Contains 21 references. (Author/PVD)

Endorsement of the money ethic and the Protestant work ethic among past and present welfare recipients was examined in a study of three groups: 164 current recipients of Aid to Families with Dependent Children (AFDC); 159 individuals receiving AFDC while enrolled in various training programs; and 158 past AFDC recipients who are now employed. The 30-item Money Ethic Scale and Blood's 4-item scale measuring endorsement of the Protestant work ethic were administered to all three groups. Demographic data were also collected. The responses to both scales were subjected to exploratory factor analyses. Four factors determined money attitudes: good, success, evil, and budget. Employed welfare recipients tended to have higher incomes, longer job tenures, and more positive attitudes toward money than did welfare recipients regardless of whether they were in training. Most welfare recipients did not value money; rather, they took money for granted and considered it evil. Employed past welfare recipients no longer consider money evil. Welfare recipients in training programs had the strongest belief that money spent is money lost (wasted). As expected, employed past welfare recipients endorsed the Protestant work ethic more highly than the other two groups did. (Contains 49 references.) (MN)

This study examined Mexican American college students' perceptions of career-related barriers and the relationship between perceptions and the students' sense of control over and responsibility for career decision making. Mexican American students (57 men and 68 women) at a state-supported university in the Southwest completed a demographic questionnaire and two career assessments: the Career Beliefs Inventory and the Career Decision-Making Self-Efficacy Scale (CDMSE)--Short Form. No sex differences among the criterion variables were identified. No significant relationship was found between the number of past career-related barriers and CDMSE; however, a significant negative relationship was found between CDMSE and the number of future barriers perceived. The more future career-related barriers students perceived, the lower their CDMSE was likely to be. Study skills and financial problems were the two most cited past and future career-related barriers. Students who exhibited relatively high levels of control and responsibility regarding career decision making tended to perceive the fewest barriers to achieving their career goals. It was recommended that counselors help Mexican American students differentiate between real and perceived barriers and develop realistic perceptions related to occupational goals and career development. (Contains 38 references.) (MN)

What the Illinois Goal Assessment Program (IGAP) test actually tests and the consequences of these tests for funding decisions were studied with a random sample of 100 school districts in the Cook County suburbs of Chicago. Eighth-grade IGAP scores for reading were obtained from the state report card, a document prepared by each school district under legislative mandate. Per pupil expenditure, attendance rate, mobility rate, average teacher salary, percentage of low income students, and the ratio of the number of students in the district to the number of teachers in the district were studied for significant correlations. Partial correlations were then used to isolate particular relationships, and analysis of variance was used to provide information for explaining variations in scores. Results support the conclusion that the statewide test in Illinois, the IGAP, measures more than student achievement. The bell-shaped curve of eighth-grade reading scores and the high and highly significant intercorrelations among all IGAP test results strongly imply the IGAP is a test of ability. Multiple regression shows that nearly three-fourths of the variation on IGAP test scores is due to context factors and not academic achievement. As the IGAP test exists, it is to a large measure a stronger indicator of poverty and mobility rate than of achievement. To a lesser extent, it is an indicator of the ratio of students to teachers, attendance rates, and cost variables. Implications for policy formation are discussed. An appendix presents two examples of test content--sixth and eighth grade reading tests--and associated questions. (Contains 12 tables and 20 references.) (SLD)

Predicted and actual scores on subtests of the Wechsler Adult Intelligence Scale-Revised (WAIS-R) and the Wide Range Achievement Test-Revised (WRAT-R) were obtained from 40 normally achieving postsecondary students (NA) and 92 adult students with learning disabilities (LD). LD students had major deficits in reading-related areas. It was hypothesized that, in their perceived deficit areas, students with LD would have different response patterns than NA students; that is, they would tend to have accurate estimates or underestimates of their responses, while NA students would tend to overestimate their performance on various tasks, as predicted by the work of A. Bandura (1989). It was also predicted that in their nonperceived deficit areas, students with LD would tend to have response patterns similar to those of NA students. Repeated measures analysis of variance found significant three-way interactions by group for predicted/actual scores by subtests for the WRAT-R and the WAIS-R. In the Reading subtest of the WRAT-R and in three of the WAIS-R verbal subtests, the first hypothesis was supported unambiguously. Students with LD had accurate predictions, while NA students overestimated significantly. In the Performance section of the WAIS-R, the second hypothesis was also supported. Differences between predicted and actual scores were similar for both groups. Implications for research and practice are discussed. (Contains 2 figures, 3 tables, and 19 references.) (Author/SLD)

Although an analysis of covariance (ANCOVA) allows for the removal of an uncontrolled source of variation that is represented by the covariates, this "correction," which occurs with the dependent variable scores is unfortunately seen by some as a blanket adjustment device that can be used with an inadequate amount of consideration for the homogeneity of slopes assumption. When regression slopes are found not to be parallel, treatment effects will most likely be biased, and there will be a reduction in the efficiency of the analysis. Twenty heuristic data sets coupled with analysis of variance and ANCOVA analyses are provided to illustrate what may occur when the homogeneity of slopes requirement is not met. Even though each of the groups had identical means, variations in the distribution of data for one of the groups studies led to varying slopes. Consequently, three different ANCOVA values resulted, only one of which was accurate. It should be noted that the homogeneity of slopes assumption can be violated to some degree without seriously affecting the robustness of tests of significance in ANCOVA. (Contains 6 tables, 23 figures, and 3 references.) (Author/SLD)

When analysis of variance (ANOVA) or linear regression is used, results may only indicate statistical significance. This statistical significance tells the researcher very little about the data being analyzed. Additional analyses need to be used to extract all the possible information obtained from a study. While a priori and post hoc comparisons can be done with qualitative data, trend analysis is most often recommended for studies with quantitative variables with fixed intervals or effects. This paper illustrates the use of trend analysis using ANOVA and multiple regression using heuristic examples. Limitations to trend analyses are also discussed. While trend analyses can be hand calculated for simple one-way ANOVAs, it is easier and more efficient to use statistical programs such as the Statistical Package for the Social Sciences (SPSS) for more complex designs such as factorials. Three appendixes illustrate trend analysis using ANOVA, regression, and multiple analysis of variance in the SPSS. (Contains 2 figures, 6 tables, and 25 references.) (Author/SLD)

Factorial analyses differ from nonfactorial analyses in that in the former all possible hypotheses (all possible main effects and interaction effects) are tested regardless of their substantive interest to the researcher and their interpretability, while in the latter, only substantive and interpretable hypotheses are tested. This paper shows the circumstances under which nonfactorial analyses are more appropriate than factorial ones. Balanced factorial analyses maximally inflate the experimentwise error rate, and in so doing, they increase the likelihood of making Type I errors. Hypothetical experiments are used to make the discussion more concrete. It is argued that only substantive and interpretable hypotheses in the design should be tested. Uninterpretable hypotheses increase the probability of making Type I error by increasing the experimentwise error rate. (Contains 2 tables and 14 references.) (Author/SLD)

This paper explains the underlying assumptions of the sampling distribution and its role in significance testing. To compute statistical significance, estimates of population parameters must be obtained so that only one sampling distribution is defined. A sampling distribution is the underlying distribution of a statistic. Sampling distributions are theoretical distributions that comprise an infinite number of sample statistics from an infinite number of randomly selected samples of a specified sample size. The influence that a large sample size has on statistical significance is demonstrated through some "what if" analyses. A "what if" analysis is simply an analysis of variance summary table in which the sample size is changed to see how statistical significance is affected. A large enough sample size invariably leads to statistical significance. Researchers with large sample sizes should look for other ways to interpret their results. One such way is effect size, which is a variance accounted for statistic that can tell how much of the variability in a dependent variable can be explained by the independent variables. (Contains 2 tables, 4 figures, and 12 references.) (SLD)

The capacity for immediate, or almost immediate, interaction between teacher and student is often cited by distance education program developers as one of the key reasons why interactive satellite courses can be as effective as face-to-face instruction. This paper analyzes the interaction between the learner and the instructor in the distance learning context by examining three research questions: (1) How important is it to provide live broadcasts, rather than taped instruction? (2) What is the importance of the instructor providing a social presence in producing learner satisfaction? and (3) What is the importance of providing interaction between the learners and the on-screen instructor? Data was collected from two Star Schools programs, one in science called "Geonauts," produced by Northern Arizona University, and one in foreign language instruction called "Elementary German," produced by Oklahoma State University. Both deliver instruction to primary/elementary schools (grades 1-6) across the United States using interactive satellite television technology. Case Studies and mail surveys of students and teachers were used to collect data. Findings are presented in four sections, dealing with live versus taped format, interaction, social presence, and a final section examining a multivariate analysis of all of these variables. The results indicate that watching the programs live had the same effect as watching them on tape. Recommendations are made for reshaping paradigms in future distance education programs. (Contains 12 references.) (AEF)

The calculation of the F statistic for a one-factor analysis of variance (ANOVA) and the construction of an ANOVA tables are easily implemented on a spreadsheet. This paper describes how to compute the p-value (observed significance level) for a particular F statistic on a spreadsheet. Decision making on a spreadsheet and applications to the classroom are also discussed for Lotus 1-2-3 for DOS or for Windows. Spreadsheets enable the learner to see what is calculated and how results are obtained, and they allow the user to automate statistical decisions. Spreadsheet use is particularly useful for students in that it creates conditions in which: (1) fundamental concepts and their meanings must be understood; (2) calculations can be automated; (3) meanings of the concepts are enhanced; (4) technology interference is minimal; (5) many examples can be studied; (6) decision making becomes the focus of learning hypothesis testing; (7) decisions about hypothesis testing can be made in different ways; and (8) changing the values on a worksheet recalculates the entire worksheet with results adjusted automatically. An appendix discusses a particular function of the Lotus 1-2-3 software for Windows. (Contains three tables and seven references.) (SLD)

This study investigates the effects of computer-aided videodisc-based anchored instruction on promoting elementary school students' problem-solving skills in Taiwan. Anchored instruction combines theories such as situated cognition, cognitive apprenticeship, cooperative learning, and constructivist theories. With the help of interactive videodisc and computer technology, anchored instruction presents situations of daily life in the way of storytelling by providing an inquiry and real-life learning environment and authentic tasks that help students enhance their problem-solving skills. Fifth-graders (n=37) from an elementary school in suburban Taipei (Taiwan) were randomly selected and divided into six groups according to their mathematical and science abilities, with two high-ability, two middle-ability, and two low-ability groups. Students were given a pretest and post-test to measure different problem-solving strategies used before and after the video instruction. Students engaged in problem solving in small groups, and were given the tests individually. Results from a two-way repeated measure analysis of variance show that student problem-solving skills improved significantly with anchored instruction. Anchored instruction provided a more motivating environment which enhanced all students' problem-solving skills, regardless of students' mathematic and science abilities. (Contains 15 references.) (SWC)

A study examined the perceptions of appropriateness regarding four types of touch between college students and professors. Subjects included 382 students, 108 men and 202 women, enrolled in a basic speech communication course at a large midwestern university. Using video depictions, an analysis of variance was conducted exploring differences in responses using gender and dyadic combinations. The four categories of touch were support, compliance, attention-getting, and affection. The arm region around the elbow was the area for the placement of touch used in the study. Results indicated that differences exist between male and female respondents based on the type of touch being used and, when compared with the other categories, the support touch was viewed as the most appropriate by both males and females. Males found the attention-getting and the compliance touches as less appropriate than females did. Both males and females found the affection touch as the least appropriate. Also identified were differences in whether touches originated from a male or female professor to a male or female student. Findings suggest that differences do exist. Future research could examine different relationships i.e.,, other personal or professional relationships. Different categories of touches could also be examined, and more contextual study on the issue of touch is needed. (Contains 41 references and 2 tables of data.) (Author/CR)

A comparative study examined the extent of growth in overall writing quality of college freshmen to determine whether feedback from the instructor during planning and the composing process (especially on the computer) was more effective than another instructor's feedback during conferences on each final, graded writing product in the students' outcomes. Subjects were 40 college freshmen, 20 in each comparison group. The study used a qualitative, pretest/posttest quasi-experimental design, with statistical analysis of variance (ANOVA). Results indicated that the students' growth in overall writing quality, between their pretest and posttest essays, was statistically significant for both comparison groups of writers. However, there was no significant difference between the two comparison groups in terms of the students' posttest outcomes in overall writing quality, although the group of students receiving feedback during their planning and composing did show slightly more growth in writing than did students receiving feedback after each of their essays had been graded. Findings suggest that both professors' personal feedback on their students' writing, combined with these instructors' methods for teaching freshman composition, made a statistically significant difference in their students' growth in their overall writing quality. (Contains 15 references; a description of essay scoring procedure is appended.) (Author/CR)

This study compared the effectiveness of Georgia community leadership programs based on leadership skill development to that of programs based on issue discussion and networking. The Leadership Practices Inventory (LPI-Self) was adapted for community leaders and mailed to a stratified sample of 376 alumni of community leadership programs based on the two approaches. Of the alumni, 188 were from rural areas and 188 were from urban areas. The 189 surveys returned were analyzed. Means and standard deviations were calculated for each of the subscales of the LPI-Self. Two-way analysis of variance was used to compare the LPI-Self subscales of alumni from the skill-based and issue- and networking-based programs. No statistically significant differences emerged between the leadership practices of the graduates of the two types of programs. Some small differences were found in only four categories of the LPI-Self subscores. Both types of programs were deemed effective in promoting the use of transformational leadership practices, and both types of curricula were concluded to be suitable for all individuals of all ages, educational levels, occupations, races, and locations (urban and rural). It was recommended that more diversity be created in community leadership classes. (MN)