【发布时间】：2023-01-31 18:46:03
【问题描述】：
拟合 LMM 后，我使用 emmeans() 函数提取估计的边际均值、SE 和置信区间。但是，取决于我是直接提取均值，还是将估计值保存为数据框，它们的 SE 和置信区间会有所不同。任何见解将不胜感激。
示例（由于字符限制无法使用 dput 和提供原始数据）：
> summary(model)
Linear mixed model fit by maximum likelihood . t-tests use Satterthwaite's method ['lmerModLmerTest']
Formula: asin(sqrt(r_index)) ~ year + prov_season + factor_month + group + prov_season * year * group + (1 | individual)
直接提取emmeans：
mm <- emmeans(model, pairwise ~ prov_season*year | group, at = list(year = c(1:8))) # extract estimates, sems and and CIs
> print(mm$emmeans)
group = naive:
prov_season year emmean SE df lower.CL upper.CL
in 1 0.0112 0.1587 309 -0.3011 0.324
off 1 0.0872 0.1768 378 -0.2604 0.435
in 2 0.0229 0.1437 253 -0.2600 0.306
off 2 0.1186 0.1577 313 -0.1916 0.429
in 3 0.0345 0.1305 203 -0.2228 0.292
off 3 0.1500 0.1405 247 -0.1268 0.427
in 4 0.0461 0.1199 162 -0.1906 0.283
off 4 0.1814 0.1261 189 -0.0674 0.430
in 5 0.0577 0.1125 136 -0.1647 0.280
off 5 0.2128 0.1155 148 -0.0155 0.441
in 6 0.0693 0.1090 125 -0.1465 0.285
off 6 0.2442 0.1098 128 0.0268 0.462
in 7 0.0810 0.1099 129 -0.1364 0.298
off 7 0.2756 0.1098 129 0.0584 0.493
in 8 0.0926 0.1149 149 -0.1345 0.320
off 8 0.3070 0.1154 151 0.0790 0.535
group = provisioned:
prov_season year emmean SE df lower.CL upper.CL
in 1 0.4076 0.0924 314 0.2258 0.589
off 1 0.2519 0.1043 413 0.0469 0.457
in 2 0.4422 0.0907 307 0.2638 0.621
off 2 0.2528 0.1000 381 0.0561 0.449
in 3 0.4768 0.0899 305 0.2999 0.654
off 3 0.2538 0.0970 355 0.0630 0.444
in 4 0.5114 0.0902 308 0.3339 0.689
off 4 0.2547 0.0952 337 0.0674 0.442
in 5 0.5461 0.0915 315 0.3659 0.726
off 5 0.2557 0.0949 329 0.0690 0.442
in 6 0.5807 0.0938 325 0.3961 0.765
off 6 0.2566 0.0959 331 0.0680 0.445
in 7 0.6153 0.0970 339 0.4245 0.806
off 7 0.2576 0.0983 342 0.0643 0.451
in 8 0.6499 0.1010 355 0.4512 0.849
off 8 0.2585 0.1019 361 0.0581 0.459
Results are averaged over the levels of: factor_month
Degrees-of-freedom method: kenward-roger
Results are given on the asin(sqrt(mu)) (not the response) scale.
Confidence level used: 0.95
emmeans as.data.frame() 的提取：
> as.data.frame(mm)
group prov_season year contrast emmean SE df lower.CL upper.CL
naive in 1 . 0.011232 0.15872 309 -0.5897 0.61217
naive off 1 . 0.087219 0.17677 378 -0.5806 0.75500
naive in 2 . 0.022854 0.14365 253 -0.5225 0.56821
naive off 2 . 0.118613 0.15767 313 -0.4783 0.71550
naive in 3 . 0.034476 0.13049 203 -0.4628 0.53172
naive off 3 . 0.150007 0.14053 247 -0.3837 0.68374
naive in 4 . 0.046098 0.11986 162 -0.4128 0.50498
naive off 4 . 0.181401 0.12615 189 -0.2999 0.66275
naive in 5 . 0.057720 0.11249 136 -0.3749 0.49036
naive off 5 . 0.212795 0.11555 148 -0.2306 0.65616
naive in 6 . 0.069342 0.10904 125 -0.3511 0.48977
naive off 6 . 0.244189 0.10984 128 -0.1790 0.66738
naive in 7 . 0.080964 0.10988 129 -0.3423 0.50419
naive off 7 . 0.275583 0.10979 129 -0.1473 0.69850
naive in 8 . 0.092586 0.11491 149 -0.3483 0.53345
naive off 8 . 0.306977 0.11541 151 -0.1356 0.74957
provisioned in 1 . 0.407628 0.09240 314 0.0578 0.75742
provisioned off 1 . 0.251854 0.10425 413 -0.1416 0.64535
provisioned in 2 . 0.442235 0.09067 307 0.0989 0.78555
provisioned off 2 . 0.252805 0.10002 381 -0.1250 0.63063
provisioned in 3 . 0.476842 0.08994 305 0.1363 0.81742
provisioned off 3 . 0.253756 0.09698 355 -0.1128 0.62035
provisioned in 4 . 0.511450 0.09023 308 0.1698 0.85310
provisioned off 4 . 0.254708 0.09524 337 -0.1055 0.61493
provisioned in 5 . 0.546057 0.09154 315 0.1995 0.89257
provisioned off 5 . 0.255659 0.09488 329 -0.1033 0.61460
provisioned in 6 . 0.580665 0.09382 325 0.2257 0.93566
provisioned off 6 . 0.256610 0.09590 331 -0.1062 0.61941
provisioned in 7 . 0.615272 0.09700 339 0.2484 0.98214
provisioned off 7 . 0.257561 0.09827 342 -0.1141 0.62920
provisioned in 8 . 0.649879 0.10101 355 0.2681 1.03170
provisioned off 8 . 0.258513 0.10190 361 -0.1266 0.64363
【问题讨论】：
-
请按照 r 标签页顶部的指南提供最少的可重现代码和输入。
标签： r linear-regression emmeans lmertest