2.6. Data analysis
Data were assessed for the assumptions of parametric statistics
(normality and homogeneity of variances, using Levene’s test).
When data did not conform to these assumptions it was transformed
by using the transformation that best corrected the detected
deviations. When no transformations yielded satisfactory
results non-parametric statistics were used. Following these
procedure:
(1) Data for both androgens (T and KT) were log transformed
and subsequently used in a repeated measures ANOVA to
test for the main effect of treatment (categorical variable:
castrated vs. sham operated vs. urinary bladder damaged)
on hormone levels in the experimental period (repeated factor:
day 0 vs. day 4 vs. day 8); planned comparisons defined
a priori were also used to assess: (A) the temporal effect (day
0 vs. day 4 and day 0 vs. day 8) of each treatment (castrated,
sham operated, urinary bladder damaged) on hormone levels;
and (B) the difference between treatments (castrated
vs. sham operated vs. urinary bladder damaged) at each
sampling point (day 0, day 4 and day 8).
(2) Behavioral data measured by instantaneous sampling (day
1–day 8) did not conform to the assumptions of parametric
statistics and therefore non-parametric tests were used;
main effect of treatment on the behavioral variables was
tested using Kruskall–Wallis tests, followed by post hoc
Mann–Whitney tests to contrast each pair of treatments
(i.e. sham vs. castration, sham vs. bladder damage, and castration
vs. bladder damage).
(3) Behavioral data measured by continuous recording (day 9)
fit parametric assumptions (except for % of time spent in display
which was log transformed) and non-transformed raw
data was used; a one-way ANOVA was used to test for the
effect of treatment (castrated vs. sham operated vs. urinary
bladder damaged) on aggressive behavior towards a putative
intruder male.