Una alternativa de intervención educativas
2.3 Adaptaciones significativas del currículo
All statistical analyses were carried out with SPSS vs 22–23 (Petrie and Watson 2006, Metsämuuronen 2009).
4.4.1 Study I
The factors affecting footpad lesions were analysed with linear mixed models for repeated measurements. Due to different numbers of observation points for different variables, two models were employed. In both models mean footpad lesion scores and the percentage of severe lesions (scores 3 and 4) were outcome variables. The first model included age and mean litter score. The second, aiming to explain further the specific effect of all factors, included the measures of litter quality, as well as bird density and slat percentage as independent variables. As litter score is expected to be an outcome of these more detailed factors, it was not included in the model.
The effect of bird age at sampling and sampling location on litter quality was analysed using separate linear mixed models for each litter quality measure. The litter quality measures, i.e. litter height, pH, moisture, and ammonia content were used as outcome variables. Age and sampling location were used as independent variables and the models included house as a random effect.
4.4.2 Study II
Litter material and platform treatment
The effects of farm and treatment (litter materials or platform) on mean footpad scores and the severity of footpad dermatitis assessed with both scoring systems, mean hock burn and mean cleanliness scores, and the distribution of hock burn and cleanliness scores were analysed with general linear univariate models for each of the dependent variables separately. Wood shavings and ground straw were compared
52
with their controls (peat) in separate models. The models included farm and litter material, and farm and platform treatment as fixed factors. The interaction between farm and litter material was included in the initial model, but omitted from the final model when the interaction was not significant. Only significant interactions are reported.
Because the data for litter condition and quality did not meet the assumptions of normality effects for litter material, platform treatment, farm, sampling time and sampling location, litter condition and quality were analysed using nonparametric tests. Effects of litter material and platform treatment, as well as of farm, on litter condition and quality were analysed with the independent samples Mann-Whitney U-test and Kruskal-Wallis test, respectively. The changes in litter height, moisture and pH over time were analysed with the Wilcoxon signed rank test. The analyses of sampling location effects on litter condition and quality were carried out using the related samples Friedman’s test and further adjusted with Bonferroni-corrections. All 16 control batches with peat were analysed together for analyses of sampling location.
Risk factors for lesions
Separate mixed models were used to identify risk factors for footpad lesions and hock burns. Mean flock footpad and hock skin lesion scores were outcome variables. The models included farm, flock, litter material, sampling location and the measures of litter quality at the end of production period (i.e. litter height, pH and moisture), as well as bird density and house size as independent variables. Factors having correlation coefficients greater than 0.6 were not introduced into the model. Thus ammonia and body weight were not included in the initial model. No correlations between other continuous independent variables were over 0.6 when detected using a Pearson correlation test. Non-significant variables were stepwise removed. Model fit was monitored by using Akaike information criteria. The normality of residuals was graphically assessed to ensure model suitability. Results of the final model are reported.
4.4.3 Study III
The use of platforms and perches was analysed using non-parametric tests as the sample size was rather limited, and the data on the use of platforms and perches were not normally distributed. The mean number of birds on perches and platforms across the 6 observations was used as the outcome variable. The development in the use of elevated structures (data pooled across day and night) by age was analysed with related samples Friedman tests separately for platform and perch use. The difference in the use of platforms and use of perches, separately, during day and night was analysed individually for each observation ages with Friedman tests.
The difference in the use of low and high perches, and thinner and thicker perches was analysed with related samples Friedman tests. The mean number of birds on perches of different height and thickness, and at different age, was analysed
53
with separate tests. The mean number of birds per single perch was used in the analyses.
4.4.4 Study IV
Due to the low use of perches, the houses with perches were excluded when analysing the difference from control houses on walking ability and TD. Effects of scoring age, treatment (platform or control) and farm on mean gait score and the percentage of birds scored 3 and 4-5 were analysed using separate general linear univariate models. For analysis, scoring age was categorized as younger (29-31 days of age) and older (34-36 days of age). The models included scoring age, treatment and farm as fixed factors.
Effects of house size, mean bird density, mean live weight at slaughter, and mean wheat percentage in feed on mean gait score and the percentage of birds scored as 3 and 4-5 were analysed using general linear multivariate model. For analysis, house size was categorized as smaller ≤ 500 m² and larger > 500 m², mean bird density as
lower ≤ 39 kg/m² and higher > 39 kg/m², mean live weight at slaughter as lower ≤ 2.41 kg and higher > 2.41 kg, and mean wheat percentage in feed as lower ≤ 19%
and higher > 19%. The model included house size, mean bird density, mean live weight at slaughter age and mean wheat percentage as fixed factors.
Since the data for TD did not meet the assumptions of normality they were analysed with nonparametric tests. TD percentage, the number of birds with each TD score assessed at slaughter, TD percentage and the number of birds with each TD score in birds with gait score 3 were used as outcome variables. The effect of platforms on TD was analysed separately using the independent samples Mann- Whitney U-test. The effect of gait scoring age on TD percentage in birds with gait score 3 was analysed with the Mann-Whitney U-test. The effect of farm on the same outcomes was analysed separately with the independent samples Kruskal-Wallis test. Data for each bird were included into the analysis according to the highest TD score. Furthermore, the effects of house size, bird density, mean live weight at slaughter, and mean wheat percentage in feed on TD were analysed separately with Mann- Whitney U-tests.
54
5 Results
Results are expressed as means and standard errors (SE) unless otherwise indicated. Statistically significant differences are reported. A P-value < 0.05 was considered statistically significant. For exact P-values, refer to papers and manuscripts at the end of the thesis.