Table 3. Results of the backward selection for the models describing the bleaching degree and mortality rates.
| Response | Groups | Variables | N | df | LRT | Pr (>Chi) |
|---|---|---|---|---|---|---|
| Bleaching degree (2016 & 2017)* | All Acropora | “Colony Size” | 121 | 1 | 0.8 | 0.38 |
| “Group” | 3 | 6.8 | 0.08 | |||
| “Time” | 1 | 68.2 | <0.00001 | |||
| Bleaching degree (2017)* | “Group” | 59 | 3 | 3.0 | 0.38 | |
| All Acropora | “Growth” | 1 | 7.0 | 0.01 | ||
| “Colony Size” | 1 | 1.0 | 0.32 | |||
| A. digitifera | “Growth” | 25 | 1 | 0.4 | 0.53 | |
| “Colony Size” | 1 | 9.3 | 0.002 | |||
| A. gemmifera | “Growth” | 23 | 1 | 0.5 | 0.49 | |
| “Colony Size” | 1 | 0.1 | 0.75 | |||
| Tabular Acropora | “Growth” | 9 | 1 | 16.9 | 0.00004 | |
| “Colony Size” | 1 | 0.00 | 1 | |||
| Partial mortality | All Acropora | “Colony Size” × “Time” × “Group” | 113 | 3 | 5.2 | 0.16 |
| “Group” × “Time” | 3 | 2.3 | 0.51 | |||
| “Colony Size” × “Group” | 3 | 5.7 | 0.13 | |||
| “Colony Size” × “Time” | 1 | 18.2 | 0.0004 | |||
| “Group” | 3 | 5.6 | 0.02 | |||
| Whole mortality | All Acropora | “Colony Size” × “Time” × “Group” | 193 | 3 | 0.7 | 0.87 |
| “Colony Size” × “Group” | 3 | 2.3 | 0.51 | |||
| “Group” × “Time” | 3 | 3.1 | 0.38 | |||
| “Colony Size” × “Time” | 1 | 1.3 | 0.26 | |||
| “Colony Size” | 1 | 0.03 | 0.87 | |||
| “Time” | 196 | 1 | 3.9 | 0.05 | ||
| “Group” | 3 | 12.3 | 0.01 |
Note:
Interactions between the explanatory variables were not included in the bleaching degree analyses because they caused technical issues (complete separation) in the model.
The variables selected in the final models have significant P values and are indicated in bold. Two models, with and without the terms of interest, were compared using a log-likelihood ratio test. If the P values were >0.05, the terms were eliminated from the model. N is the number of observations for each model. df is the degree of freedom, which equals to the difference in the number of parameters between the models being compared. LRT is the log-likelihood ratio based on the difference in the residual deviance of the two models. Significant P values (<0.05) are indicated in italics and bold.