Statistical modelling

Study A

There are now alternative designs for dealing with experiments in which large numbers of plant varieties need to be compared. These designs are known as row-column designs which can efficiently handle experimental spatial variation in both directions. The late Mr Harvey Dicks provided the following solution for 54 (53+1) accessions in a 3-replicated 18x3 row-column design which he presented during a training course in Gaborone in June 2009 in which this case study was discussed.

The numbers from 1-54 correspond to 54 accessions, one more than the 53 tested in the original experiment. Should an additional accession not be available then one of the 53 could have been duplicated. (Naturally a design with the prime number 53 is not possible.)

Note that each accession occurs once in each replicate and the columns are 'latinised' in the sense that no one accession occurs twice in any one column. This is clearly a more efficient design than the conventional randomised block since variation can be accounted for in two directions.