The object of this study is to present improvements on a method of analysis of field experiments, which were presented in 1937. These improvements might be summarized as follows:
1) By designing hypothetical experiments on uniformity trials, we can determine "real" error, and compare "experimentally" different methods; in this way I have proved, that my method is more accurate than the classical ones, which are moreover inaccurate in the determination of the error. 2) In the field there is correlation between the productivity of nearby plots, and accuracy of whatever method depends on the advantage it takes of such correlation; correlation is maximum between adjacent plots, and that is why my method is the more accurate. 3) In the field there are gaps (sudden changes of productivity) between adjacent plots, and abnormal plots, whose productivity differs greatly from both adjacent and their average; the method permits to determine them objectively and increase accuracy. 4) Yield adjustment can be also done on the basis of edaphic and other criteria, and such adjustment can be combined with that on adjacent plots, this is promising. 5) To compute error, the paper proposes a "simulation" method, which simulates the way, in which productivity deviations between plots compensate really one another, when treatment averages are computed; it reduces considerably the difference between computed and "real" error. 6) Low or negative correlation between adjacent plots is usually due to competition between border lines, long ago known as "inteference", such competition distorts yield differences between treatments, and makes the experiment useless: measures are taken to minimize its effect; in the case of tree experiments it is very frequent, and can be avoided by separating plots with ditches sufficiently deep, to avoid the passage of roots from one plot to another. 7) For a long time analysis of field experiments has been based on purely theoretical considerations; but spatial distribution of soil heterogeneity is also important; combination of the two aspects may be useful.
Το 1937 δημοσιεύθηκε μια μέθοδος ανάλυσης των πειραμάτων στον αγρό, η οποία αυξάνει σημαντικά την ακρίβεια του πειράματος. Ύστερα από 30 χρόνια εγκατάλειψης της μεθόδου, μετά τον τελευταίο παγκόσμιο πόλεμο, η μέθοδος τράβηξε τα τελευταία χρόνια την προσοχή. Σκοπός της μελέτης αυτής είναι να καταστούν γνωστές οι τελειοποιήσεις της μεθόδου και να συζητηθούν μερικές απόψεις του θέματος που αγνοούνται.
