The experimental set-up was a randomised complete block
design with four replicates. Data were analysed within years and
replicates were regarded as a random factor. Analyses of the main
factors of soil, seed treatment, the interaction between soil and
seed treatment and replicates were performed byF-tests. Means of
main factors were separated by the least significant difference
(LSD) and were declared different at theP¼0.05 probability level.
The analyses were performed using the PROC GLM module within
the SAS/STAT®software package 9.2 (SAS Institute Inc., Cary, NC,
USA)