This research be proposed evaluate the statistical significance and importance of
sources of variation (SV) in GxE interaction studies in the cotton (Gossypium
hirsutum L.) crop for cotton-seed yield (RENDAS), fiber percentage (POFIB) and
fiber yield (RENDIF), in the Colombian dry Caribbean (CS) and humid Caribbean
(CH). Four data sets were used, which were obtained from the evaluations of ten
different genotypes of medium fiber in agronomic evaluation tests (PEA), in a
completely randomized block design (CRBD) with four replications. The data were
taken from the 2003/2004 (CS and CH), 2007/2008 (CS) and 2009/2010 (CH) cotton
crops, in four environments representing the CS and CH producing areas. For each
of the three response variables included in the study, a combined analysis of
variance was performed for each data set, from the four environments of each
production zone, assuming a mixed model with genotypes (G) as fixed effects and
environments (E) as random effects. With the mean squares (MS) of the combined
analysis of variance (ANAVACO), the statistical significance was determined and
with the percentage of the sum of squares (SS) the importance of SV. The results
showed that most of the variation in SS for RENDAS in CS was associated with E,
which presented a highly significant difference (p<0,01), representing an average of
90,1%, followed by GxE with 7,4% and G with 2,6%. In CH there was a similar trend,
with an explanation for the variation in E of 71,8%, followed by GxE with 17,1% and
G with 11,2%. For POFIB, most of the variability in CS was associated with G, which
presented a highly significant difference, with an average of 74,4%, followed by E
with 18%, also with a highly significant difference, and 7,6% for GxE. In CH, E
obtained 57,2% of the variation, followed by G with 34,7% and GxE with 8,1%, all
three sources being significant. The greatest contribution to the SS of RENDIF in the
CS was given by E, which presented a highly significant difference and explained on
average 87,2% of the variation, followed by GxE with 8,1% and G with 4,7%. In CH,
the effect of E, also presented a highly significant difference and responded with
64% of the variation of SS, followed by G with 18,2% and GxE with 17,9%. The
results of the PEA in the cotton crop in CS and CH showed similar trends in the
statistical significance and in the explanation of the variation of SS, highlighting the
effects of E, as the most important, but with higher values in CS than CH, and the
variation due to E had more effect in the expression of RENDAS and RENDIF.
Therefore, it is suggested to increase the number of trials per PEA in more than four
environments in the Colombian Caribbean.
Keywords: cotton GxE interaction studies, sources of variation, mean square, sum
of square, ANAVACO, Colombian Caribbean.