TABLE 2.
Results of mixed models of studied recruitment traits (A) germination percentage, (B) mean time to germination (MTG), (C) emergence percentage, (D) mean time to emergence (MTE), (E) survival percentage after the growing season “gs,” and (F) overwinter survival percentage (“ow”), in response to distribution, temperature conditions, the interaction term, and ecological indicator values (see Table 1) that were found to be significant when individually added to the models.
| Model terms | Full model (m1) | Chisq (m1) | p-value (m1) |
| (A) Germination percentage | Log-Odds: Est. (95% CI) | ||
| Distribution | 0.05 (−1.63, 1.72) | 0.12 | 0.732 |
| Temperature | 0.68 (0.59, 0.77) | 378.72 | <0.001 |
| Eco-value W | 1.60 (0.69, 2.51) | 9.42 | <0.01 |
| Distr × Temp° | 0.51 (0.31, 0.71) | 26.64 | <0.001 |
| (B) MTG | Est. (95% CI) | ||
| Distribution | −3.77 (−9.39, 1.85) | 3.15 | 0.076 |
| Temperature | −4.30 (−5.46, −3.15) | 46.02 | <0.001 |
| Eco-value W | −6.56 (−8.82, −4.30) | 16.81 | <0.001 |
| Distr × Temp° | −2.05 (−5.00, 0.90) | 1.85 | 0.173 |
| (C) Emergence percentage | Log-Odds: Est. (95% CI) | ||
| Distribution | −0.14 (−0.75, 0.47) | 0.21 | 0.646 |
| Eco-value H | −0.43 (−0.72, −0.15) | 7.70 | <0.01 |
| Eco-value LF (h/t) | 0.14 (−0.65, 0.92)/−2.71 (−3.78, −1.65) | 17.96 | <0.001 |
| (D) MTE | Est. (95% CI) | ||
| Distribution | −0.32 (−0.92, 0.28) | 1.10 | 0.295 |
| Eco-value T | 0.61 (0.26, 0.96) | 9.35 | <0.01 |
| Eco-value H | 0.49 (0.22, 0.75) | 9.32 | <0.01 |
| (E) Survival percentage “gs” | Log-Odds: Est. (95% CI) | ||
| Distribution | 0.62 (−1.23, 1.48) | 1.85 | 0.174 |
| Year garden (2017) | −2.70 (−3.16, −2.24) | 80.08 | <0.001 |
| (F) Survival percentage “ow” | Log-Odds: Est. (95% CI) | ||
| Distribution | −0.25 (−0.50, 0.00) | 3.48 | 0.062 |
| Eco-value F | −0.38 (−0.51, −0.25) | 11.25 | <0.001 |
All models were set with distribution “wide” as reference, temperature “warm” in cases (A,B), LF “c” in case of (C). Est., Estimate; CI, Confidence intervals; Chisq, test statistic and p-values of full model (m1). Chi-square and p-values were estimated with maximum likelihood; °in cases of interactions we used contrast-coding (Levy, 2018).