The authors contend that most research in education lacks statistical power. They feel that the poor use of statistics as a tool for enhancing internal validity must be remediated. The adoption of a new convention is proposed in order to put statistical certainty into reasonable perspective. (RLV)

This paper presents an empirical analysis of similarities and differences between two statistics, G and Phi, which treat genuinely dichotomous data. These results can aid researchers in selecting between these two statistics and in evaluating results from the use of one v the other. (Author)

It has been shown that the degree of restriction of range taken into account in testing the hypothesis that rho equals zero, entails risks of incorrect inferences. It is argued that an alternative is to disregard the restriction of range and to use the common t-statistics proposed by regression theory. (Author/JKS)

Two applications of Kristof's theorem on traces of matrix products are presented in order to highlight their utility for psychometric theory and studies. (Author/JKS)

Social scientists have available a wide array of techniques for the analysis of complex contingency tables. Considerations in choosing from among the techniques are discussed. The theme emphasized here is that the analyst should evaluate the techniques within the context of the substantive questions of the study. (Author/JKS)

Illustrates how expected mean squares needed in the analysis of variance can be arrived at via the use of only one rule: the expected mean square E(MS) for any source of variation for any ANOVA model is specified effect plus the specified effect in combination with any random effect. (Editor/RK)

The main findings were that, in its present form, the Staff Morale Questionnaire is not suitable to measure teacher morale in England, that there is a need for future research into teacher morale, and that older teachers apparently display higher morale than younger teachers. (Author)

Estimates of the internal consistency and reliability of the first principal component are provided through the use of the largest characteristic root and associated vector of the equicorrelation matrix. The estimate of the internal consistency is also shown to be a lower bound for the measure provided by Horn (1969). (Author)

Compares the methods used to measure the educational attainment of respondents to the Current Population Survey and the decennial census. Describes a way to reconcile old and new methods of determining highest grade completed. (JOW)

Considers pluses and minuses of establishing a corporate style guide. Surveys articles and conference presentations (published between 1985 and 1995) on style guides. Analyzes significance and contribution of each. Suggests areas for further research, specifically quantitative analysis of the value of style guides. (PA)

Explains that Type I error is a risk undertaken whenever significance tests are conducted, and the chances of committing a Type I error increase as the number of significance tests increases. Notes that adjusting the alpha level because of the number of tests conducted in a given study has no principled basis, commits one to absurd beliefs and…

Tries to distinguish the key insights from some distortions by clarifying the goals of experiment-wise error control that D. O'Keefe correctly identifies as vague and open to misuse. Concludes that a better understanding of the goal of experiment-wise error correction erases many of these "absurdities," but the clarifications necessary…

Responds to Daniel O'Keefe's "Against Familywise Alpha Adjustment," where O'Keefe maintains that one should never attempt to control Type I error introduced when many statistical tests are conducted. Argues that alpha adjustment should be applied only in the narrowly circumscribed instance when the researcher wants to make a strong claim…

Responds to other articles in this colloquy. Notes that they propose that alpha adjustment can appropriately be mandated when the set of tests concerns the key claims of a single theory. Considers how consistent application of this reasoning commits one to bizarre beliefs and research practices--which suggests that this reasoning does not provide…

Neural network models and more conventional statistics were used to study the use of sequential phonological probabilities in the segmentation of a phonological transcription of conversational English that might allow the infant to bootstrap into a series of increasingly sophisticated segmentation competencies. Results are interpreted in the…