血球計算器内の赤血球分布に關する研究(赤血球算定の誤差に關する研究―後編)

抄録

1) For re-examining of the problem of homogeneous distribution of red corpuscles in a counting cell, each datum observed with a Bürker-Türk's chamber was tested in two ways as follows.
First, corpuscles in 5 large squares of ruled area were counted in 480 rectangle unitareas, one of which was fourfold the size of a small square. The homogeneity of corpuscles distribution in these 5 large squares was tested “statistically”, i, e. by the χ² test of the discrepancy between the frequency distribution of these 480 data and the expected Poisson distribution of equal mean (Test I). On the other hand, the corpuscles in each large square were counted in 16 middle sized squares, one of which was sixteenfold the size of a small square. From the 5 large squares was counted consequently a datum consisting of 5 sets of 16 values. Concerning this datum the homogeneity of corpuscles-distribution among the 5 large squares way tested “Stochastically”, i, e. by a test of “homogeneityamong the means of small
samples” (Test II). As far as the distribution of red corpuscles in a countingcell is theoretically homogeneous, the data of every counting case must be “satisfactory” in the goodness of fit at Test I and “not significant” at Test II. It was found, however, in the counting with routine procedures, that one out of 2 cases which were satisfactory at Test I was significant at Test II, and that 8 out of 24 independently tried cases were significant at Test II.
2) It was found by the following experiments that the main cause of such an uneven distribution was due to the procedure of charging a counting cell.
a) Concernig the manner of charging a counting cell with diluted fluid, the experiments were divided in 2 groups: charging at one sitting (C. I . S.) and charging at two sitting or more (C. II. S.). Among 11 cases of C. II. S, 5 cases of uneven distribution in test II were observed, while among 11 cases of C. I . S., none of such cases was observed. The difference between the uneven distribution cases of two groups was significant with 0.05 level.
b) No case of the uneven distribution in Test II was observed among 18 cases of C. I . S. in which the counting cells were not filled completely with fluid, while few of such uneven distribution cases were observed among this group C. I . S, cases in which the spaces were overfilled, even if slightly, with an excess of fuid.
c) The counting was made with the same pipet and with the same cell on 20 blood samples (Experiment A). The Experiment A was divided in two groups of ten cases. Group I was made with paying no special regard to a charging conditon. Group II was made under two special charging conditions: charging the counting cell at one sitting, and with no excess of fluid. Among 10 cases of group I, 3 cases of uneven distribution in Test II were observed and each F- value (ratio between interclass variation and interclass variation) of all cases deviated remarkably from zero-point, while among 10 cases of group II, no case of such uneven distribution was observed and each F-value did not deviate significantly.
3) Five countig celles of Bürker-Türkes' type were filled successively with diluted fluid of a pipet, in accordance with the above mentioned charging procedure, and the corpuscles contained in each central large square of these cells were counted in 16 middle sized squares. The data consisting of 5 sets of 16 values counted were tested similarly as in Test II (Experiment B). This was divided in 2 groups. Group I of Experiment B consisted of 5 cases being made with paying no special regard to a charging condition and group if of the same experiment consisted of 10 cases being made under the above mentioned special charging conditions. One case of the uneven distribution was observed in group I of Experiment B, while