我已经采用线性混合模型进行了分块图设计,以评估水,氮和磷对BWC(生物量加权的2c值,通过将每个物种的2C值的乘积相加得出(DNA含量)及其生物量分数(物种子图生物量/总子图生物量):
model1.1<-lmer(log(BWC)~W*N*P+(1|year)+(1|W:Block),data=BWC)
W(0,1),N(0,1)和p(0,1)有两个级别,我想使用boxplot将线性混合模型的输出报告给我。但是,我对线性混合模型的输出感到困惑。
在model1.1中W N P的估计值(斜率)为负,这是否意味着WNP处理与对照图相比将减少BWC?但是我们可以看到,在WNP处理下,箱中BWC最高。
summary()和anova()之间存在差异,例如,N和P效果的重要性。 N的估计值为-4.0911,这意味着N的添加减少了BWC,但是N的影响不明显。如何报告类似N的治疗效果?
非常感谢您的任何评论。 在BWC上进行WNP治疗的箱线图: enter image description here
https://i.stack.imgur.com/cKOFt.png (很抱歉,这些链接似乎至少需要10个信誉才能发布图像)
summary()和anova()输出:
> summary(model1)
Linear mixed model fit by REML. t-tests use Satterthwaite's method ['lmerModLmerTest']
Formula: BWC ~ W * N * P + (1 | year) + (1 | W:Block)
Data: BWC
REML criterion at convergence: 2969.1
Scaled residuals:
Min 1Q Median 3Q Max
-2.93847 -0.71228 -0.07573 0.68191 2.92589
Random effects:
Groups Name Variance Std.Dev.
W:Block (Intercept) 0.9169 0.9575
year (Intercept) 0.8346 0.9136
Residual 18.2966 4.2774
Number of obs: 515, groups: W:Block, 14; year, 10
Fixed effects:
Estimate Std. Error df t value Pr(>|t|)
(Intercept) 10.8498 0.6985 46.5200 15.532 < 2e-16 ***
W1 2.0844 0.8969 45.9613 2.324 0.02460 *
N1 -4.0911 0.7364 486.0288 -5.556 4.56e-08 ***
P1 -2.0460 0.7600 490.1120 -2.692 0.00734 **
W1:N1 4.6738 1.0394 485.9800 4.497 8.65e-06 ***
W1:P1 0.9695 1.0687 485.9809 0.907 0.36478
N1:P1 5.7550 1.0687 485.9773 5.385 1.13e-07 ***
W1:N1:P1 -3.3306 1.5100 485.9541 -2.206 0.02788 *
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Correlation of Fixed Effects:
(Intr) W1 N1 P1 W1:N1 W1:P1 N1:P1
W1 -0.645
N1 -0.531 0.414
P1 -0.515 0.401 0.488
W1:N1 0.376 -0.582 -0.708 -0.346
W1:P1 0.366 -0.566 -0.347 -0.706 0.488
N1:P1 0.366 -0.285 -0.689 -0.706 0.488 0.502
W1:N1:P1 -0.259 0.400 0.488 0.499 -0.688 -0.708 -0.708
> anova(model1)
Type III Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
W 750.15 750.15 1 11.90 40.9995 3.519e-05 ***
N 10.84 10.84 1 485.95 0.5926 0.44177
P 29.14 29.14 1 494.92 1.5926 0.20755
W:N 290.51 290.51 1 485.95 15.8778 7.793e-05 ***
W:P 15.54 15.54 1 485.96 0.8493 0.35721
N:P 536.85 536.85 1 485.95 29.3415 9.562e-08 ***
W:N:P 89.01 89.01 1 485.95 4.8648 0.02788 *
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
> emmeans::emmeans(model1,pairwise~N*P*W)
$emmeans
N P W emmean SE df lower.CL upper.CL
0 0 0 10.85 0.699 46.9 9.44 12.26
1 0 0 6.76 0.696 46.2 5.36 8.16
0 1 0 8.80 0.721 52.1 7.36 10.25
1 1 0 10.47 0.721 52.1 9.02 11.91
0 0 1 12.93 0.696 46.2 11.53 14.33
1 0 1 13.52 0.696 46.2 12.12 14.92
0 1 1 11.86 0.721 52.1 10.41 13.30
1 1 1 14.86 0.721 52.1 13.42 16.31
Degrees-of-freedom method: kenward-roger
Confidence level used: 0.95