Ellis B. Page

Ordered Hypotheses for Multiple Treatments: A Significance Test for Linear Ranks

Abstract In many experiments the background evidence, theories, or conditions suggest an expected ordering among the treatment effects, yet in the analysis of variance such implicit hypotheses are typically neglected. A ranking statistic L is presented as test of a monotonic relationship among the treatment groups in the two-way analysis of variance. Used with accompanying table of L, it combines considerable power with computational ease, and assumes data of only ordinal strength. L is related to the test of the linear component of the treatment sum of squares in the parametric randomized-block design, to the product-moment correlation and regression, to the normal deviate test of Lyerlys average rho, and to Friedmans chi-square of ranks. Where either L or the Friedman test may be used, L is often more accurate and appropriate, and it has some advantages over other tests of trend and monotonicity.

Computer Grading of Student Prose, Using Modern Concepts and Software

Abstract In earlier work of Project Essay Grade (PEG) we used computers to evaluate prose of high school students. In major experiments, PEG successfully imitated single human ratings, despite the crude hardware and software of the late 1960s. Today, computers are common in home and school, and advanced software packages permit much more powerful analysis. In the present research we analyzed recent federal samples of 495 and 599 essays and simulated groups of human judges, reaching multiple Rs as high as .87, close to the apparent reliability of the targeted judge groups. We also generated weights from formative samples of two thirds each, which predicted well the other one-third samples (with simple rs higher than .84). Another cross-validation predicted across different years, students, and judge panels, with an r of .83. Thus, the computer surpassed two judges, which is the usual human panel. Results appear encouraging for further research and indeed for early application to large programs of essay eva...

Family Configuration and Mental Ability: Two Theories Contrasted with U. S. Data

Recent articles have argued that family size and birth order have important cognitive influence, and may partially explain the different performances of social class and ethnic groups. Support for such claims has often depended on the analytic strategy, which has concentrated on the means of aggregated groups with large N’s. For the present analysis, data are reviewed from the massive U. S. National Longitudinal Study of Educational Effects. Results from aggregate analysis are quite similar to those reported by Zajonc and others. When individual variation is explored, however, the effects of family configuration become relatively trivial, and the confluence theory appears untenable. The apparent effects of family size, far from explaining population differences, seem themselves to be better explained as the result of group admixtures. And the small, residual birth-order effects therefore appear to result from other phenomena, still to be explained.

Effects of U.S. Private Schools: A Technical Analysis of Two Recent Claims

Private High Schools Are Better Than Public, Study Concludes A major study by sociologist James S. Coleman concludes that Catholic and other private high schools provide a better education than public ones do, and in some respects are less racially segregated. ... [Supporters of public schools] fear that the findings could strengthen the case for tax credits to families paying tuition for private schools. (pp. Al, A13)

Family Size, Birth Order, and Intelligence in a Large South American Sample

The most popular current theory of family influence is the confluence theory, which hypothesizes that a child is helped or hindered in intellectual development according to the average absolute intelligence (mental age) in the family when that child is born. Aggregated data about such a hypothesis have been examined in previous studies from a number of nations, but not from a developing country. Here test scores, family information, and socioeconomic data are analyzed far a sample of over 36,000 college applicants in Colombia. The intellectual effects of family size were not at all the classic pattern: All family sizes smaller than six surpassed a single child family, arguing that differences were populational, rather than intrafamilial. Further analysis showed almost no family-size effect for the lower socioeconomic group among the college applicants, and birth order effects were not constant across family sizes, and not in conformity with the model. In sum, the confluence theory fared badly in these tests.

Do Catholic High Schools Improve Minority Student Achievement

There is much current debate concerning the role of private schooling in U. S. education. Recent research has purported to show that Catholic schools produce higher achievement in minority high school seniors than do public schools, yet this research has failed to control adequately for student ability, frequently a criterion for selection into such schools. Here, the High School and Beyond data set and path analytic techniques were used to compare black and Hispanic high school seniors’ achievement in public and in Catholic schools. When better measures of ability were added to the causal models, the apparent effect of Catholic schooling on minority achievement was greatly reduced from the former claims. Yet a small, meaningful path still remained. Further analyses suggest that any such Catholic school advantage may be due, in part, to their more stringent curriculum. In any case, the apparent advantage exists only for minority students, and the claimed overall achievement advantage remains unproved.

Trait Ratings for Automated Essay Grading

This study employed an automated grader to evaluate essays, both holistically and with the rating of traits (content, organization, style, mechanics, and creativity) for Webbased student essays serving as placement tests at a large Midwestern university. The authors report the results of two combined experiments, based on random selection from 1,193 essays. In the first experiment, the essays of 807 students were used to create statistical predictions for the essay-grading software. In the second experiment, the ratings from a separate, random sample of 386 essays were used to compare the ratings of six human judges against those generated by the computer. The interjudge correlation of the human raters alone was r = .71. But the interrater reliability of all six judges in combination with computer scoring reached .83. The essay-grading software was an efficient means for evaluating the essays, with a capacity for grading approximately six documents every second. Other potential feedback measures for use in writing courses are also discussed.

Massive Intervention and Child Intelligence: The Milwaukee Project in Critical Perspective

The Milwaukee Project, conducted by R. Heber, H. Garber, and others with small numbers of inner-city preschool children, has been described as raising IQs from the dull-normal to the superior ranges. Despite its fame, however, it has rarely been studied technically or critically. An earlier (1972) writing by Page reached the conclusion that the project was characterized by biased selection of treatment groups, contamination of criterion tests, failure to specify the treatments, and inaccessibility of data to professional analysis. This newer article reviews the current status of these problems and presents information that has become available since 1972. The long-promised final report has not been issued, but what evidence there is suggests a decline of the experimental, massive-intervention children to the level of the untreated controls in those measures, such as school reading, which are not under the control of the project.

Statistical and linguistic strategies in the computer grading of essays

Essay tests are used in the schools and colleges of all nations, and in major testing programs of national and eveh international size. Potentially, such essay tests are an important applied field for computational linguistics, and.should eventually provide focus for much work. Yet in the past, little direct attention has been paid to such grading, although there are ways to begin investigation which would not necessarily require much linguistic knowledge beyond that now available.

Tests and Decisions for the Handicapped: a Guide to Evaluation Under the New Laws

1Portions of this monograph are based on a position paper that was produced under Contract No. 300-77-0237 with the Department of Health, Education, and Welfare, US Office of Education, Bureau of Education for the Handicapped, and is in the public domain. Opinions expressed herein do not necessarily reflect the position or policy of the Bureau, Office of Education, or the Department, and no official endorsement should be inferred. P.L. 94-142 and the guidelines