ミニコンによる非線形最小二乗あてはめ計算の一改良案

抄録

Non-linear least square method has become an indispensable procedure for the analysis of pharmacokinetic data. This method, however, occasionally gives much bigger sum of square value to make the computation chaotic. To overcome this difficulty, various procedures have been devised. In this connection, after deliberation of characteristic features of normal equation and its coefficient matrix, "temporal discard" procedure was designed. When failed to give smaller sum of square value, the row and the column of the coefficient matrix which includes the biggest or the smallest element on the principal diagonal are discarded. The search for the least square parameters is continued with smaller normal matrix thus obtained. Preliminary test calculation gave satisfactory results.