インフルエンザウイルスの一定量法としてのLi-Rivers培養法の意義について 第2報

ウイルスの感染価の推定とその信頼限界及び2つの測定値の差の検定について

抄録

The author, assuming that the individual sensitivity of experimental animal is normally distributed against the logarithmic dosis of infective viral agent, proposed methods for determining fifty per cent endpoint and its confidence limit and for testing the significance of the difference of two fifty per cent endpoints, and applied them to the experiments of influenza virus cultivated with the method of Li and Rivers.
A normal regression line, which was estimated most likely from the values of experimental probits of the ratio of the reaction at various doses of virus, was adopted as an approximation of the linear relationship between the logarithmic dosis of virus and the value of probit of corresponding population. Fifty per cent endpoint was estimated as an intersection of virus axis and regression line. Confidence limit was estimated as intersections of virus axis and envelopes of regression lines which correspond to the points of borderline of the confidence ellipse of the two variables - gradient coefficient of the regression line and the intersection of probit axis and regression linc. Lastly a method for testing the significance of difference of two fifty per cent endpoints was given, by assuming that, though the fifty per cent endpoints are not equal, the gradient coefficients of the two different regression lines are equal since both of them were concerned with the same strain of virus and with the same method of cultivation.