Ardie Lubin

The Interpretation of Significant Interaction

on certain varieties of potatoes. Even then he realized that the usual additive model, where each observed score equals the sum of the main effects, might be inadequate in many cases and tried a product formula. Nowadays an interaction term for each potato-manure combination would be inserted to see whether each manure effect is the same for all potato varieties. When the interaction term is significant, the additive model has failed; we cannot obtain unbiased estimates of the treatment effects. The purpose of this article is to discuss what can be done about the

SOME EFFECTS OF SLEEP LOSS ON MEMORY

Immediate recall of word lists showed significant impairment after one night of sleep loss. Since S was required to write down each word immediately after its presentation, the deficit was not due to failure of sensory registration. With 24-hr. delayed testing, a picture-recognition test did not show significant deficit after one night of sleep loss. Performance on this test was impaired, however, after a night of recovery sleep. These results imply that moderate sleep loss causes deficit in formation of the memory trace rather than in storage or retrieval functions and that this effect is probably independent of the physiological lapses (brief periods of sleep) which affect vigilance and sensory registration.

EEG frequency and finger pulse volume as predictors of reaction time during sleep loss

Abstract The validity of EEG frequency and finger pulse volume for predicting reaction time increased as sleep loss increased. In general, the EEG showed the highest correlation with reaction time, especially in the 1 sec interval just before and just after the signal. As sleep loss increased, the “lead time” for the EEG increased so that by 50 h of sleep loss, valid predictions of reaction time could be made at least 2–3 sec before the signal. Five out of seven subjects showed a bimodal EEG frequency distribution during sleep loss. For these five subjects, perceptual-motor lapses occurred during periods in which the EEG frequency was in the theta rhythm (4–7 c/sec) range. For two of the subjects whose modal EEG frequency slowed somewhat, but did not have a secondary mode at 4–7 c/sec, neither the EEG nor finger pulse volume could be used to predict reaction time. Finger vasodilation was significantly related to long reaction times during sleep loss, but its contribution to the prediction of reaction time was small.

The Effect of Correlation On the Repeated Measures Design

Abstract : The purpose of this paper is to point out that: (a) any repeated measurements on the same organism will in general exhibit statistical dependence; therefore multivariate analysis of variance rather than univariate analysis of variance is appropriate, and (b) multivariate analysis of variance assumes that the carryover effect of a treatment or a test on succeeding treatments is constant and does not depend on the nature of the succeeding treatment, i.e., carryover is additive and does not interact with succeeding treatments. Most of this paper is concerned with possible experimental and statistical answers to the questions which arise when dependent measures are used in a continuous treatment design. The problem of carryover effects that interact with subsequent treatments is quite different. No answers to this problem are given here; instead we ask if there is, in fact, any way of preserving the advantages of a cross-over design and obtaining unbiased estimates of the treatment effects when carry-over interaction is present.

A theory of pattern analysis for the prediction of a quantitative criterion

A method of pattern analysis is presented for the case of dichotomous items and a quantitative criterion. This “configural scale” has maximum validity in the least squares sense. A technique for computing the configural scale as a polynomial function of the item scores is given. Tests of significance are outlined for such questions as: Is there a linear or non-linear relation between the quantitative criterion and the item scores? Does the addition of certain items to the test increase the validity of the configural scale? Are all the items in the configural scale fully effective?

The use of configural analysis for the evaluation of test scoring methods

A method based on configural analysis has been given whereby test scoring techniques can be evaluated to see if they have optimal validity. Configural analysis has also been used to show how three well known item scoring techniques, multiple regression, total score, and multiple cut-off, imply (for optimal validity) certain conditions on the answer pattern means. The method is illustrated by a worked example.

The Use of Configural Analysis for the Prediction of a Qualitative Criterion

HORST (1954) in an analysis of Meehl’s paradox (1950) first showed how a polynomial equation could be used for the prediction of a dichotomous criterion, from two dichotomous items, in the case where a linear combination of the items had little or no validity. In an extension of Horst’s results Alf (1956) developed a multiple phi coefficient to express the correlation between a configural score based on k items, and a dichotomous criterion. Lubin and Osburn (1957) and Osburn and Lubin (1957) presented a general polynomial equation for the optimal prediction of a quantitative criterion from k dichotomous items. The method was called configural analysis following the terminology of Meehl (1950). McQuitty (1957) in a discussion of some of the issues in pattern analysis has emphasized the desirability of extending configural analysis to the case of a polychotomous criterion. As McQuitty points out, Lubin and Osburn (1955) indicated some time ago that their results on the prediction of a quantitative criterion can be readily extended to the qualitative case. The purpose of this paper is to derive configural equations appropriate for the case of k dichotomous items and a polychotomous cri-