John D. Morris

Multivariate analysis versus multiple univariate analyses.

The argument for preceding multiple analysis of variance (ANOVAS) with a multivariate analysis of variance (MANOVA) to control for Type I error is challenged. Several situations are discussed in which multiple ANOVAS might be conducted without the necessity of a preliminary MANOVA. Three reasons for considering a multivariate analysis are discussed: to identify outcome variable system constructs, to select variable subsets, and to determine variable relative worth. The analyses discussed in this article are those appropriate in research situations in which analysis of variance techniques are useful. These analyses are used to study the effects of treatment variables on outcome/response variables (in ex post facto as well as experimental studies). We speak of an univariate analysis of variance (ANOVA) when a single outcome variable is involved; when multiple outcome variables are involved, it is a multivariate analysis of variance (MANOVA). (Covariance analyses may also be included.) With multiple outcome variables, the typical analysis approach used in the group-comparison context, at least in the behavioral sciences, is to either (a) conduct multiple ANOVAs or (b) conduct a MANOVA followed by multiple ANOVAS. That these are two popular choices may be concluded from a survey of some prominent behavioral science journals. The 1986 issues

The Effect of Note Taking on Ninth Grade Students' Comprehension.

The effect of note-taking instruction on ninth graders’ comprehension of highand low-interest passages on teacher-made, objective tests was explored. A sample of 115 World Cultures students, aged 13–15 (48 females. 67 males; 6 Blacks, 109 Whites) from a suburban junior high school participated. The treatment group (N = 61) received 9 weeks of instruction and practice in the Cornell method of note taking, and the control group (N = 54) did not. The effects of treatment, ability level, and passage type (low-interest or high-interest) were investigated. A 2 × 2 MANOVA revealed a statistically significant main effect for note-taking training, F (2, 110) 5.88, p < .01. In addition, a statistically significant interaction, F(1, 111) = 7.57, p < .005, between note-taking training and passage type suggested that the training was more effective for the low-interest than the high-interest passage. There was no statistically significant interaction between ability level and passage type.

Validation of the Lollipop Test: A Diagnostic Screening Test of School Readiness:

The validity of the Lollipop Test: A Diagnostic Screening Test of School Readiness was examined by using the Metropolitan Readiness Test (MRT), Level I, Form Q, as the criterion. The sample of 293 kindergarten pupils was administered the MRT by their teachers in classroom groups; the Lollipop Test was individually administered by qualified examiners. The statistical significance of all correlations (p < .001) demonstrated appreciable concurrent validity across the test batteries. Further, a canonical correlation indicated a high degree of multivariate relationship between the tests. Implications of these results were discussed with respect to school readiness screening and the use of the Lollipop Test.

A Single Contrast Test Procedure

The thesis of this article is that the multitude of procedures for testing hypotheses about mean contrasts often presented in statistical methods textbooks is seemingly unwarranted. Nearly all research situations calling for the study of contrasts can be handled with a single contrast test statistic, that often attributed to R. A. Fisher. By jointly considering a probability value and an eta-squared value and by keeping in mind the total number of contrasts studied, one finds that a single contrast test procedure evolves.

A Comparison of Regression Prediction Accuracy on Several Types of Factor Scores

Several advantages to the use of factor scores as independent variables in a multiple regression equation have been advocated in the literature. To provide guidance for selecting the most desirable type of factor score upon which to calculate a regression equation, computer-based Monte Carlo methods were used to compare the predictive accuracy upon replication of regression on five “complete” and four “incomplete” factor score estimation methods. For several levels of multiple correlation (R2 = .30, .50, and. 70), and for several subject-to-variable sampling ratios (3:1, 5:1, and 10:1), prediction on incomplete factor scores showed better double cross-validated prediction accuracy than on complete factor scores. Moreover, the unique unit-weighted factor score was superior among the incomplete methods.

Predicting Later Academic Achievement from Kindergarten Scores on the Metropolitan Readiness Tests and the Lollipop Test

The long-term predictive validities of the Metropolitan Readiness Tests (MRT) and the Lollipop Test: A Diagnostic Screening Test of School Readiness were examined. The achievement of 246 students in reading and mathematics as measured by the Stanford Achievement Test and teacher assigned grades in first, third, and fourth grades was predicted from kindergarten administrations of each of these test batteries. All multiple correlations for the Lollipop Test and the MRT were found to be significant and similar in magnitude. Perhaps particularly noteworthy, was that the Lollipop Test, a shorter screening instrument, performed as well as the lengthier MRT in predicting school achievement.

Behavior ratings of emotionally disturbed children by teachers, parents, and school psychologists

Emotional disturbance was rated by teachers, parents, and the school psychologist for a group of 104 children diagnosed as severely emotionally disturbed. Teachers were found significantly more severe in their judgments than the school psychologist on the Behavior and Socialization scales of the Referral Checklist, but not on Communication. The teacher rating profile was also found to deviate from parallelness from the other raters with increased severity in the Behavior scale. None of the nine correlations between judges on the same scales was significant. Interpretations and possible implications are offered.

Variables associated with teacher retention: A multi‐year study

Abstract This research study is a multi‐year project covering four South Florida School Districts that examines variables associated with the demographics of 2,129 teachers who were hired during the 2000–2001 academic year. The purpose was to assess the relationship between teacher retention and various demographic variables. A model predicting teacher retention based upon these variables is also intended. Findings revealed that 96% of the sample were retained after 1 year of employment, 79% were retained after 2 years and 72% after 3 years. The variables of gender, χ2(1) = 13.03,p < .007, out‐of‐state hires, χ2(1) = 15.17,p < .007, preparation, χ2(2) = 9.93, p < .007, and placement, χ2(1) = 7.81, p < .007, were significantly associated with retention after the 3rd year of the study.

THE WISC-R SUBTEST PROFILE OF A SAMPLE OF SEVERELY EMOTIONALLY DISTURBED CHILDREN

The WISC-R subtest profiles of 113 children classified as severely emotionally disturbed (88 males and 25 females; 71 Caucasians and 42 Negroes) ranging in age from 6 yr., 11 mo. to 13 yr., 8 mo. was examined. Diagnosis was based on psychological testing and quantitative assessment of behavioral deviations by parents, teachers, and psychologists. Scores for Caucasian children were significantly superior to those of Negro children, on the Information, Similarities, Vocabulary, and Picture Arrangement subtests. However, all subtest means foe both races were significantly lower than those in the standardization sample. A multivariate test of interaction and a Hotelling T2 suggested that the profiles of Caucasian and Negro subjects were not “flat” as those in the standardization sample and also not parallel. Investigation of the shape of the profiles of the races showed an elevated Picture Completion score for the Negro children and a depressed Coding score for Caucasian children. No evidence supported a discrepancy between Verbal and Performance abilities.

The Superiority of Factor Scores as Predictors.

Common factor scores were compared to unfactored data-level variables as predictors in terms of the correlation of a criterion with the predicted value in multiple regression equations applied to replication (cross-validation) samples. Data were generated by computer to provide populations with three different degrees of common variance inherent in their predictor variable intercorrelation matrices. Two replication populations differing from the original by specified amounts in their intercorrelation matrices were created for each common variance level. Results indicated that shrinkage was less for factor scores than for data-level variables for all combinations of common variance and difference of replication population. Moreover, the actual correlation describing accuracy of prediction was higher for factor scores than for data-level variables at the extreme conditions of common variance and difference of replication population.