[Statistical study of results of Hemalog D (author's transl)]

The data which Hemalog D obtains from blood samples of non hematologic patients have been analyzed statistically and compared with those obtained from an eye count. As well as usual statistical techniques, i.e. correlation coefficient and paired t-test, a more sophisticated multidimensional method was used : the Factorial Analysis of Correspondences. We have analyzed two samples. For the first sample, that of "all-comers", the reproductibility of the results given by the machine was shown to be excellent, and always superior to that of the eye count. The machinen-technician agreement is reasonably good except for the monocytes and basophiles. The preceeding conclusions were found to be true for the second sample, i.e. "with discrepancies", and in addition, it was shown that an eye count on 400 elements gives a result closer to that of the machine than does a reading on 100 elements.