*2.2. Statistical Analysis*

Descriptive statistics of the data collected were obtained using an online statistical software (Statulatorbeta®, Sydney, Australia) with year, season, month, species, category, country of provenance, nationality of transport company and country of destination as categorical variables, while NATT, DOA, DCP, total dead, UFT and space allowance were used as numerical variables.

Chi-square tests were conducted to determine the association between factors (i.e., year, season, month, species, typology, provenance, destination). Trucks with DOA, DCP, and/or UFT were considered as having a welfare problem. A univariate logistic regression model was developed with welfare problem as a binary outcome (1/0; welfare problem/non-welfare problem) and year, season, month, species, category, provenance, destination, nationality, and space allowance as predictive variables. *P* values were calculated using the Wald test. Each predictor variable returning a *p* < 0.25 from the univariate analyses was considered for inclusion in the final multivariate model for welfare

problems. Predictor variables for the final multivariate logistic regression model were selected using a step-wise backward elimination procedure, whereby predictive variables were removed until all variables in the final model had a *p* < 0.05 indicating significance [9]. The aforementioned statistical analyses were performed using GenStat®Version 14 (VSN International, Hemel Hempstead, UK).

The effect of year, season, month, provenance, destination, transport company nationality, species and category on NATT was determined using a General Linear Model (GLM) procedure. Tukey's HSD (honestly significant difference) test was used as a post-hoc test. Statistical analyses were performed using SAS version 9.4. *P* threshold was set at 0.05. Data are expressed as least square means ± standard error (SE).
