我希望有人能通过我的解释确认我在正确的轨道上。我已经使用序数包中的clmm拟合了混合效果序数模型,我想知道在响应变量(sc_1_concern)的每个类别中的预测概率。 要使用emmeans到达那里,是否正确:
mmod_sc1_con_fin <- clmm(sc_1_concern ~ visit + randomization_group + (1|id), data = dat)
# Emmeans
mmod_sc1_con_fin_em <- emmeans(mmod_sc1_con_fin, ~ sc_1_concern + visit + randomization_group, mode = "prob")
> mmod_sc1_con_fin_em
sc_1_concern visit randomization_group prob SE df asymp.LCL asymp.UCL
1 0 1 0.002447062 0.0012377990 Inf 2.102043e-05 0.004873103
2 0 1 0.188293704 0.0326846079 Inf 1.242330e-01 0.252354358
3 0 1 0.710193915 0.0359065593 Inf 6.398184e-01 0.780569479
4 0 1 0.099065319 0.0222581619 Inf 5.544012e-02 0.142690514
1 1 1 0.001646361 0.0008720446 Inf -6.281519e-05 0.003355537
2 1 1 0.135129929 0.0288390201 Inf 7.860649e-02 0.191653370
3 1 1 0.722649649 0.0350093036 Inf 6.540327e-01 0.791266623
4 1 1 0.140574061 0.0314491529 Inf 7.893485e-02 0.202213268
1 0 2 0.002244366 0.0011549238 Inf -1.924253e-05 0.004507976
2 0 2 0.175475588 0.0318478870 Inf 1.130549e-01 0.237896299
3 0 2 0.715206223 0.0358373487 Inf 6.449663e-01 0.785446136
4 0 2 0.107073823 0.0229827071 Inf 6.202854e-02 0.152119101
1 1 2 0.001509889 0.0008137213 Inf -8.497568e-05 0.003104753
2 1 2 0.125351953 0.0278619587 Inf 7.074352e-02 0.179960388
3 1 2 0.721763804 0.0351831932 Inf 6.528060e-01 0.790721596
4 1 2 0.151374354 0.0325782105 Inf 8.752224e-02 0.215226474
# Plot prediction
emmip(mmod_sc1_con_fin_em, sc_1_concern ~ visit|randomization_group)
我不确定我是否需要使用“ prob”或“ cum.prob”。这些比例与观察到的比例不完全匹配(它们在球场上),但我认为这与模型有关(比例奇数似乎令人满意)。谢谢