On the statistical estimation of speech-organization distributions from aphasia data

Statistical methods of estimating speech-organization distributions from aphasia data are presented. Under certain assumptions, estimators for the proportions of individuals with unilateral right-hemisphere speech, unilateral left-hemisphere speech and bilateral speech organization are derived, and the large-sample properties of these estimators are presented. It is shown that useful results can be obtained when the most questionable assumption is relaxed. The general methodology is presented in terms of a specific example--the estimation of speech organization distributions of right- and left-handed adults across the three speech types. Under the relaxed assumptions a test of the hypothesis that all left-handers are unilaterally organized is derived. Conservative bounds on the proportions of unilaterally organized (left- and right-hemisphere) and bilaterally organized left-handers are also derived. As an illustration, the results are applied to aphasia-incidence data from the literature. Several uncertainties and inadequacies of the available data are discussed. Although minor modifications of the present methods might be required for other examples, the main ideas would be the same.