我在 R 中有以下混合效应回归模型：
linearModSubset <- lmer(od.deaths.pc ~ (year.x|County.x) +
pov_perc +
unemply_perc +
hispanic_perc +
sing_parent_hh_perc +
Fentanyl.Crime.Lab.Cases.pc +
Meth.Crime.Lab.Cases.pc +
Benzo.Crime.Lab.Cases.pc +
COC.Crime.Lab.Cases.pc +
opioidHospitalizations.pc +
popDens,
data=drug.data.county.year.filt)
summary(linearModSubset)
r.squaredGLMM(linearModSubset)
tab_model(linearModSubset)
使用summary(linearModSubset)and r.squaredGLMM(linearModSubset)，我得到：
Linear mixed model fit by REML. t-tests use Satterthwaite's method ['lmerModLmerTest']
Formula: od.deaths.pc ~ (year.x | County.x) + pov_perc + unemply_perc +
hispanic_perc + sing_parent_hh_perc + Fentanyl.Crime.Lab.Cases.pc +
Meth.Crime.Lab.Cases.pc + Benzo.Crime.Lab.Cases.pc + COC.Crime.Lab.Cases.pc + opioidHospitalizations.pc + popDens
Data: drug.data.county.year.filt
REML criterion at convergence: 2142.3
Scaled residuals:
Min 1Q Median 3Q Max
-3.6337 -0.5535 -0.0570 0.3850 6.0012
Random effects:
Groups Name Variance Std.Dev. Corr
County.x (Intercept) 1.068e+02 10.334251
year.x 1.922e-05 0.004384 -1.00
Residual 1.066e+02 10.323681
Number of obs: 284, groups: County.x, 72
Fixed effects:
Estimate Std. Error df t value Pr(>|t|)
(Intercept) -2.708189 2.441282 50.613193 -1.109 0.272531
pov_perc 57.142106 22.800314 50.891021 2.506 0.015442 *
unemply_perc 222.294220 66.788253 55.417666 3.328 0.001557 **
hispanic_perc 0.452461 0.307941 41.392374 1.469 0.149309
sing_parent_hh_perc -45.872855 12.961035 221.598936 -3.539 0.000489 ***
Fentanyl.Crime.Lab.Cases.pc 0.345118 0.056659 248.968846 6.091 4.23e-09 ***
Meth.Crime.Lab.Cases.pc -0.016265 0.007532 67.392572 -2.159 0.034379 *
Benzo.Crime.Lab.Cases.pc -0.246739 0.099567 272.662523 -2.478 0.013814 *
COC.Crime.Lab.Cases.pc 0.090246 0.035264 148.696936 2.559 0.011491 *
opioidHospitalizations.pc 0.022111 0.003516 63.809163 6.289 3.26e-08 ***
popDens 0.005422 0.002347 35.207351 2.310 0.026864 *
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Correlation of Fixed Effects:
(Intr) pv_prc unmpl_ hspnc_ sng___ F.C.L. M.C.L. B.C.L. COC.C. opdHs.
pov_perc -0.379
unemply_prc -0.185 -0.217
hispanc_prc -0.286 0.089 -0.087
sng_prnt_h_ -0.309 -0.253 -0.376 -0.140
Fntn.C.L.C. -0.193 0.049 -0.169 0.019 0.210
Mth.Cr.L.C. -0.080 0.013 -0.034 0.303 -0.192 -0.077
Bnz.Cr.L.C. 0.035 -0.098 0.003 0.005 0.071 -0.119 -0.039
COC.Cr.L.C. 0.099 0.104 -0.024 -0.082 -0.209 -0.271 0.099 -0.296
opdHsptlzt. 0.343 -0.310 -0.145 0.001 -0.317 -0.178 0.012 -0.148 -0.016
popDens 0.217 -0.030 0.111 -0.553 -0.087 -0.042 -0.010 0.075 -0.013 -0.066
fit warnings:
Some predictor variables are on very different scales: consider rescaling
optimizer (nloptwrap) convergence code: 0 (OK)
boundary (singular) fit: see ?isSingular
> r.squaredGLMM(linearModSubset)
R2m R2c
[1,] 0.5414291 0.5507347
但是使用tab_model(linearModSubset)我得到：
当我使用相同的模型时，我对为什么这两个输出的值略有不同感到困惑。我也想知道为什么r.squaredGLMM(linearModSubset)报告条件 R 平方，但tab_model(linearModSubset)条件 R 平方是 NA。
提前致谢！