我目前正在进行荟萃分析，研究伐木在热带森林中的影响。
作为其中的一部分，我一直在测试效果是否因区域和使用的记录方法而异。
我使用R。
中的 metafor 包来完成所有这些操作我的数据如下：
structure(list(Method = structure(c(2L, 2L, 1L, 1L, 1L, 1L, 1L,
2L, 2L, 1L, 2L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L
), .Label = c("Conventional", "RIL"), class = "factor"), MU = c(192.96,
252.41, 235.6, 258, 258, 399, 313, 409.8, 420.4, 333.47, 327.54,
256, 228.1, 547.1, 453.3873094, 427.495, 346.8, 330.833333333333,
343.3, 221.5, 194.8, 51.1, 276), SSU = c(3, 3, 30, 3, 3, 2, 5,
17, 10, 4, 4, 4, 9, 15, 35, 10, 3, 3, 3, 3, 3, 3, 10), ML = c(157.03,
171.97, 219.5, 198, 148, 191, 204, 315.3647059, 386.22, 135.8,
211.78, 183.8, 159.9, 230.8, 97.00798294, 218.31, 279.933333333333,
261.4, 249.733333333333, 118.6, 42.9, 18.7, 128.4), SSL = c(3,
3, 10, 3, 3, 10, 5, 17, 10, 4, 4, 4, 9, 10, 131, 45, 3, 3, 3,
3, 3, 3, 10), Region = structure(c(3L, 3L, 2L, 2L, 2L, 3L, 2L,
2L, 1L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 2L, 2L, 2L, 3L, 3L, 3L, 3L
), .Label = c("Africa", "Americas", "Asia & Oceania"), class = "factor"),
SDU = c(7.69030558560582, 12.1243556529821, 74.4902678207026,
30, 30, 145, 107.33126291999, 64.9, 92.95, 40.73364703, 54.0371067,
53.6, 98.1, 193.8, 16.13693527, 109.3250955, 28.21329474,
30.91865671942, 32.220024829289, 37.065887281974, 96.4752299815865,
37.4122974434878, 91.706052144883), SDL = c(8.46972844901181,
7.81154914213564, 53.1262646908288, 18, 10, 324.8738217,
84.970583144992, 44.90907399, 109.0794186, 20.75198304, 18.6400617,
11.6, 88.2, 104.2, 4.008416039, 185.9464001, 29.85034897,
28.7292533839639, 15.297494348204, 37.7587076050015, 32.5625551822949,
7.44781847254617, 126.174878640718)), .Names = c("Method",
"MU", "SSU", "ML", "SSL", "Region", "SDU", "SDL"), row.names = c(NA,
23L), class = "data.frame")
然后我用它来计算我有数据的每个站点的效果大小和相关的SE，如下所示：
require("metafor")
ROM <- escalc(data=AGB, measure="ROM", m2i=MU, sd2i=SDU,
n2i=SSU, m1i=ML, sd1i=SDL, n1i=SSL, append=TRUE)
我的问题是我不知道如何解释具有两个分类预测因子的模型的治疗对比。
我的'最佳'模型（AIC最低的模型）看起来像这样：
ROM.ma1 <- rma(yi,vi,mods=~Method+Region,method="ML",data=ROM)
使用随机效应模型。
summary(ROM.ma1)
告诉我们：
Mixed-Effects Model (k = 23; tau^2 estimator: ML)
logLik deviance AIC BIC
-4.2852 65.8950 18.5705 24.2479
tau^2 (estimated amount of residual heterogeneity): 0.0634 (SE = 0.0241)
tau (square root of estimated tau^2 value): 0.2519
I^2 (residual heterogeneity / unaccounted variability): 90.58%
H^2 (unaccounted variability / sampling variability): 10.62
Test for Residual Heterogeneity:
QE(df = 19) = 616.2226, p-val < .0001
Test of Moderators (coefficient(s) 2,3,4):
QM(df = 3) = 17.0683, p-val = 0.0007
Model Results:
estimate se zval pval ci.lb ci.ub
intrcpt -0.4392 0.3150 -1.3944 0.1632 -1.0566 0.1781
MethodRIL 0.3544 0.1513 2.3420 0.0192 0.0578 0.6510 *
RegionAmericas 0.1027 0.3201 0.3208 0.7484 -0.5247 0.7301
RegionAsia & Oceania -0.3487 0.3068 -1.1365 0.2557 -0.9500 0.2526
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
现在我了解截距是因子Method和Region的第一级的组合。
我希望能够做的是计算每个组的系数估计值以及它们的置信区间，以便我可以绘制该测试的结果。
我有办法做到这一点吗？
我问了很多同事，但没有一个给我一个有用的回应。
提前致谢。