The optimum piat size in experinients with rubber trees.

Abstract

The mathematjcal-statjsucal study ol the problern of optimum plot sire in experimenta with rubber frees, by the methodology ol the lntraclass correlation coefficient (p), was carried out with the aid of data from four Brazilian experimenta. The methodology applied defines as optirnum plot size the number (k) ot test tiees per piot that minimizes the variance of a tieafrnent mean, for a fuxed total number (N) of tiees, per tieatinent. In the case of a complete guard row, ii Ied to recommend of 6 lo 12 test tines in two rows, or 9 to 15 in three rows. Plots with one half guard row should have 4 tu 10 test planta in two rows, or 6 tu 15 in three rows. The usual number of replications in experimenta with njbber tiees (3, in rnost cases) is not enough, in general. When small plota are used, greater is the need tu increase the number of replications, which should not be less than S. Bul even with more replications, b'ials with small plots altow a considerable reduclion in the experimental area, without loss of precision.