Systat 13

The Shapiro-Wilk test was conducted to evaluate normality. One-way analysis of variance (ANOVA) and the Kruskal-Wallis test were performed for parametric and non-parametric data, respectively. All data are expressed as means ± SEM. A p-value < 0.05 was considered significant. Systat 13.2 (Systat Software, Chicago, IL, USA) was used for all statistical analyses.

Statistical analyses were performed by using the Systat 13.2 software (Systat Software, Inc., San Jose, CA, USA) and GraphPad v6.01 (Prism) software. The X² test or Fisher’s exact test for categorical variables and the Mann–Whitney U test for continuous non-parametric variables were used. The risk of virulence genes to modify the MIC of the antibiotic was established by determining the odds ratio with a CI: 95%.

During the scoring of isoflurane signals, burst suppression was observed to occur both bihemispherically and unihemispherically. Consequently, an interhemispheric asymmetry index (L-R/L+R; L and R = 0.78–3.90 Hz power density for the left and right hemisphere of each brain region, respectively; following Rattenborg et al., 2016 (link)) was calculated for all isoflurane recordings. Briefly, EEG data were fast Fourier transformed in 0.78 Hz bins, applied to Hamming-windowed data, using REMLogic 3.4 (Embla, Broomfield, CO, United States). General linear mixed-effects models were fitted for each variable using Systat (SYSTAT 13.2; Systat Software, Inc.). Models were structured with bird identity as a random factor, isoflurane level as a fixed factor, and the asymmetry index value (arcsine square root transformed to meet the assumptions underlying the mixed model) as the dependent variable.

The count data were analyzed using the Loglinear component within SYSTAT 13.0 (Systat Software, Inc., San Jose, CA, USA), with two discrete variables, treatment and time-period. We examined the test for Model terms panel comparing the likelihood-ratio chi-square for the full model to the same value for the smaller model. To determine whether the removal of a term results in a significant decrease in the fit, we then looked at the difference in these statistics. If the fit was worsened with the removal of the term, it remained in the model. With each change in the model, we checked Raftery’s BIC (Bayesian Information Criterion), which when negative, concludes that the model is preferable to the saturated model. For each factor in the model, z-scores were provided within the analysis, and the probabilities associated with the z-scores were obtained using z-tables. The results are expressed as a treatment mean or the effect of the treatments on a measured count variable, and this is compared with the effects of individual treatments.