使用鲁棒 lmer 进行虚拟编码和对比编码后简单斜率分析的不同 p 值

Question

我想平均一些连续预测变量对结果变量的影响，所以我在这里使用了对比编码（ https://towardsdatascience.com/how-to-correctly-interpret-your-continuous-and-categorical-variable-interactions -in-regressions-51e5eed5de1e )

但是，我注意到当我运行简单的斜率分析时，p 值会发生变化（以及一些固定效应，代码部分未显示）。

为什么？ 哪个是最适合趋势分析的 rlmer model？

您可以在下面看到代码和 output（首先是虚拟编码，然后是对比编码）

非常感谢！

时间、年龄、常见问题、医疗都是数字 步态是一个因素（正常 = 参考猫，异常）

tmtar = rlmer(TMT_A ~
                GAIT_*TIME+
                AGE+
                FAQ+
                medical+(1|SITE:RID), 
              data  = df) 

trend
rrg = qdrg(formula = ~ GAIT_*TIME+AGE+FAQ+medical, data = df, coef = fixef(tmtar), vcov = as.matrix(vcov(tmtar)), at = list(TIME = c(0,36)))
update(pairs(rrg, reverse = T, simple = "TIME"), by = NULL)

output

Fixed effects:
                    Estimate Std. Error t value
(Intercept)        18.178377   7.691289   2.364
GAIT_abnormal      -0.928555   2.758425  -0.337
TIME               -0.002426   0.024642  -0.098
AGE                 0.270401   0.104385   2.590
FAQ                 0.715653   0.068665  10.422
medical            -0.027960   0.302248  -0.093
GAIT_abnormal:TIME  0.272333   0.067935   4.009

trend
 contrast GAIT_     AGE FAQ medical estimate    SE  df z.ratio p.value
 36 - 0   normal   74.8 6.5    6.52  -0.0873 0.887 Inf -0.098  0.9216 
 36 - 0   abnormal 74.8 6.5    6.52   9.7167 2.389 Inf  4.068  <.0001 

对比编码

contrasts(df$GAIT_) <- c(-0.5,0.5)
tmtar = rlmer(TMT_A ~
                GAIT_*TIME+
                AGE+
                FAQ+
                medical+(1|SITE:RID), 
              data  = df) 
rrg = qdrg(formula = ~ GAIT_*TIME+AGE+FAQ+medical, data = df, coef = fixef(tmtar), vcov = as.matrix(vcov(tmtar)), at = list(TIME = c(0,36)))
update(pairs(rrg, reverse = T, simple = "TIME"), by = NULL)

output

contrast
        [,1]
normal   -0.5
abnormal  0.5

Fixed effects:
            Estimate Std. Error t value
(Intercept) 17.71410    7.95539   2.227
GAIT_1      -0.92856    2.75843  -0.337
TIME         0.13374    0.03676   3.639
AGE          0.27040    0.10438   2.590
FAQ          0.71565    0.06866  10.422
medical     -0.02796    0.30225  -0.093
GAIT_1:TIME  0.27233    0.06794   4.009

 contrast GAIT_     AGE FAQ medical estimate   SE  df z.ratio p.value
 36 - 0   normal   74.8 6.5    6.52     4.81 1.32 Inf 3.639   0.0003 
 36 - 0   abnormal 74.8 6.5    6.52    14.62 3.56 Inf 4.110   <.0001