Question

在R

中构造具有随机误差项的非线性结构方程模型

library(piecewiseSEM)
library(lme4)
library(nlme)

head(VTmodelmeansadults)
            id Round Cycle Day Stage Temp Type   Sample Pos   md possal posabd  mat.max mat.mean adj.r8    cage matage
70 1 G1 3 27 1      1    G1   3  G1 3   27  abd 27 1 abd   0 0.25      0      0 2.426511   0.6378      0 G1 3 27      3
73 1 G1 3 27.2      1    G1   3  G1 3   27  abd 27.2 abd   0 0.25      0      0 2.426511   0.6378      0 G1 3 27      3
76 1 G1 3 27.3      1    G1   3  G1 3   27  abd 27.3 abd   0 0.25      0      0 2.426511   0.6378      0 G1 3 27      3
82 1 G1 3 27.4      1    G1   3  G1 3   27  abd 27.4 abd   0 0.25      0      0 2.426511   0.6378      0 G1 3 27      3
85 1 G1 3 27.5      1    G1   3  G1 3   27  abd 27.5 abd   0 0.25      0      0 2.426511   0.6378      0 G1 3 27      3
88 1 G1 3 27.7      1    G1   3  G1 3   27  abd 27.7 abd   0 0.25      0      0 2.426511   0.6378      0 G1 3 27      3

从数据中可以看出，我想知道day，md和cycle对二项式结果Pos的影响。 md本身受温度和成熟度的影响，并且笼子有随机效应。

 model = list(
   lme(md ~ Temp + matage, random = ~ 1|cage, na.action = na.omit,
       data = VTmodelmeansadults),

   glmer(Pos ~ Day + md + Cycle +(1|cage), 
         family=binomial(link = "logit"), data = VTmodelmeansadults))

summary(model)
     Length Class    Mode
[1,] 18     lme      list
[2,]  1     glmerMod S4

评估模型匹配时，我收到以下错误消息

sem.fit(model, VTmodelmeansadults)
  |========================================|  75%
Error in MEEM(object, conLin, control$niterEM) : 
  Singularity in backsolve at level 0, block 1

我已经尝试用lmer（）代替lme（），以防出现别名问题但是会收到此错误

model = list(
   lmer(md ~ Temp + matage + (1|cage), na.action = na.omit,
      data = VTmodelmeansadults),

  glmer(Pos ~ Day + md + Cycle +(1|cage), 
         family=binomial(link = "logit"), data = VTmodelmeansadults))

sem.fit(model, VTmodelmeansadults)
  |==================================================================|  75%
Error in KRmodcomp.lmerMod(basis.mod.new, basis.mod.drop) : 
  Models have either equal fixed mean stucture or are not nested'

它仍然是一个别名问题吗？或者glmer模型可能与lmer模型冲突？

非常感谢任何帮助。

改编自jshlefe超级有用的piecewiseSEM包的方法 https://github.com/jslefche/piecewiseSEM

R中具有随机效应误差的R结构方程模型

0 个答案: