G. A. Baker

Strawberry Uniformity Yield Trials

Realistic and comprehensive mathematical models for the analyses of field trial data can be based only on an extensive and intensive study of uniformity trials. The collection of uniformity trial data is a time and money consuming project. It cannot be assumed that the models found adequate for one crop in one year at one place can be extrapolated with any degree of confidence to other quite different crops in very different environments. This is a well recognized problem and several workers (1), (2), (3), (4), (5), (6), (7), (8), and (9) have published recent results. The purpose of the present paper is to present two uniformity yield trials for strawberries. The variety grown was the same in the two cases but the environments were somewhat different for the different trials. Some calculations are made and presented including random sampling results in order to emphasize differences between the trials and point to possible arrangements of field trials that will lead to improved clarity of interpretation.

Empiric Investigation of a Test of Homogeneity for Populations Composed of Normal Distributions

Abstract A test of homogeneity is developed that will frequently detect non-homogeneity by means of samples as small as fifty for some populations that are composed of normal components. The development is essentially the same as Smirnovs distribution free tests as discussed in [4].

Factor Analysis of High School Variables and Success in University Subjects for the First Semester in the University

A PREVIOUS study (3) has been made of the relationships between high school scholastic ex perience and success in the first semester i n the University of California at Davis. In this previous study, success in the first semester in the University was measured in four different ways all of which turned out to be practi c a 11 y equivalent. These four variables were b a sed on courses undertaken by the students during their first semester. All students did not take identical programs. If high school experience prepares for some courses better than for others then part of the failure to predict Univer sity success from high school experience com pletely is due to the fact that the programs of the students are not identical. Thus, it becomes important to study success in each subject in terms of preceding high school experience to see to what extent success is ex pressible in terms of high school performance. In this study it is found that the different com mon first semester University courses differ markedly in their profiles with respect to a set of typical high school variables, and, also, with respect to the extent to which high school vari ables determine or measure success in the var ious courses.

The effect of selection on linear functions of normally distributed correlated variables on the distributions of other linear functions

A general method of studying the effect of selection on linear functions of variables with multinormal distributions on the distribution of other linear functions has been indicated. In order to show exactly what is involved an example is discussed in considerable detail. It is seen that for practical application much numerical calculation is involved which could be partially relieved by more adequate tables and possible adaption of electronic computers.