We experimentally observed differences in susceptibility to both carbaryl and potassium dichromate among the studied pond populations of D. magna (analysis of deviance, P < 0.001).
An analysis of deviance associated with this model, with the variance scaled to allow for overdispersion, was used to determine significance of pruning in relation to the presence of N. fuckeliana.
The analysis of deviance of the logistic regression model of the total incidence of infection per tree across the course of the study indicated no significant difference in N. fuckeliana infection between pruned and unpruned trees (F1,167=0.24, p=0.62; Figure 3a).
As indicator for the improvement from the null model to the two-level model, we calculated McFadden's (1973) adjusted R-squared (= 1−[−2LL2-level model/−2LLnull model] = 1−[9,827.192−10,039.848]) and found an error reduction of only 2.1percentt, but which remains significant by analysis of deviance (χ2-test).
With this defined, an analysis of deviance using stepwise multivariate linear regression models was performed to determine if the Torg ratio was associated with canal area and if a "below normal" Torg ratio was predictive of lumbar stenosis for each subject.
Model selection was carried out using Akaike's Information Criterion (AIC) [29] and analysis of deviance.
The binary mildew infection data were analysed using multiple logistic regression with analysis of deviance [30].
In the shallowest depth zone (50 150 m), a linear model was sufficient (analysis of deviance, p = 0.12) to describe the relation between AvTD and latitude.
Analysis of deviance (ANODEV), and Pearson's chi-square or Fisher's exact tests for differences among the lineages were used, as appropriate.
When the degrees of freedom for the smoothed term were near their minimum value, we examined whether the trend was described sufficiently by a linear regression using Akaike's information criterion (AIC) and analysis of deviance between the two models.
