Probit Analysis

177 mental psychology wlherein deduced concepts and inductions from clinical observation are to be coml)ined in a comprehensive frame of reference within which psyclhological pheenomena in lhumans are to be understood. When a responise mlade by a living organism to a stimtultus must fall inlto one or the other of two mutually exclusive classes, it is knowvn as a quantal or all-or-nothing response. For example, the two classes mighlt be deatlh or survival, tuber-culini positive or tuberculin negative, pregnant or nonp)regnant. If the stimuli coInsist of several different doses of a druig, a quiantal response curve may be con-structe(d to show the p)ercent of test subjects that give a positive response at the variotus (lose levels. Thle treatment of such data falls witlhin the province of probit analysis. Objectives of the analysis miglht be the estimation of the median effective close, that is the close at wlhiclh 50% of the test subjects show a positive responise, or the estimation of the potency of some biological preparation, or drug, relative to the potency of a standard preparation. Probit analysis is not restricted to data arising in plharmacology. The age of onset of some developmental event, such as tlle erul)tion of a partictular tooth, may be studhiedI by the same methods. Subjects are classified by age ancl are furtlher grouped into two classes according to wNlietlher the event lhas or lhas niot occurred. Age is treated as the stimulus and the developmental event as the response. One can estimate the mean age at wlhich the event occurs and the stand-arcl error of this mean. Finney's book, now in its third edition, was the first to be devoted to probit analysis and has preempted the field. It is a remarkably successful work and the current revision incorporates much new material including several interesting examples and a chapter on appropriate computer methods. The first three clhapters provide an introduction to the subject for those who wish to make a simple and approximate analysis of data. A method based on a visual fit of the probit regression line is carefully described. In addition to providing a simple method for a practical problem, this chapter introduces the principle of the analysis, and so forms a starting point for more detailed discussion. Before developing this the author touches on some side issues. He does not approve of such approximate methods as the Reed-'Muench. Against the Spear-man-Karber method, …