如何解决警告 glmer：GLMM 失败
我必须运行一个带有连续响应变量的 glmer，两个变量作为固定效应和随机效应：
glmer.1 <- glmer(Conc ~ Metodo * Kit +
(1|Especie),data = Input_2,family = Gamma(link = "inverse"))
我收到此警告消息：
Warning messages:
1: In checkConv(attr(opt,"derivs"),opt$par,ctrl = control$checkConv,:
Model failed to converge with max|grad| = 0.0652105 (tol = 0.002,component 1)
2: In checkConv(attr(opt,:
Model is nearly unidentifiable: very large eigenvalue
- Rescale variables?
我使用 bobyqa() 作为优化参数进行了尝试，并收到了以下警告消息：
glmer.Conc<-glmer(Conc ~ Metodo * Kit
+ (1 | Especie),control = glmerControl(optimizer = "bobyqa"),family = Gamma)
Warning messages:
1: In checkConv(attr(opt,:
Model failed to converge with max|grad| = 0.0747655 (tol = 0.002,:
Model is nearly unidentifiable: very large eigenvalue
- Rescale variables?
我的总结如下：
summary(glmer.1)
Generalized linear mixed model fit by maximum likelihood (Laplace Approximation) ['glmerMod']
Family: Gamma ( inverse )
Formula: Conc ~ Metodo * Kit + (1 | Especie)
Data: Input_2
AIC BIC logLik deviance df.resid
3202.0 3223.6 -1595.0 3190.0 264
Scaled residuals:
Min 1Q Median 3Q Max
-0.7347 -0.5647 -0.2876 0.2740 12.0396
Random effects:
Groups Name Variance Std.Dev.
Especie (Intercept) 5.264e-05 0.007256
Residual 1.724e+00 1.313030
Number of obs: 270,groups: Especie,6
Fixed effects:
Estimate Std. Error t value Pr(>|z|)
(Intercept) 0.0089191 0.0028893 3.087 0.00202 **
Metodo -0.0002581 0.0009247 -0.279 0.78013
Kit -0.0009974 0.0007818 -1.276 0.20203
Metodo:Kit 0.0007497 0.0004608 1.627 0.10380
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Correlation of Fixed Effects:
(Intr) Metodo Kit
Metodo -0.531
Kit -0.531 0.857
Metodo:Kit 0.449 -0.907 -0.897
convergence code: 0
Model failed to converge with max|grad| = 0.0652105 (tol = 0.002,component 1)
Model is nearly unidentifiable: very large eigenvalue
- Rescale variables?
Especie 代表物种的类型，metodo 和 kit 代表使用什么类型的方法和试剂盒获得浓度，Conc 是浓度值
我的数据框如下所示：
Input_2 <- readr::read_table2("Especie Metodo Kit Conc A1 A2
1 1 1 70.4 1.5 0.9
1 1 1 57.2 1.7 1.4
1 1 1 22.2 1.6 1.1
1 1 1 60.6 1.5 0.9
1 1 1 50.2 1.6 1
1 1 2 40.5 2.1 1.7
1 1 2 86.6 2.1 2.1
1 1 2 53.9 1.8 1.3
1 1 2 50.3 1.9 0.2
1 1 2 44.1 2 1.5
1 1 3 159.1 2.1 2.3
1 1 3 246 2.1 2.4
1 1 3 81.5 2 1.3
1 1 3 107.9 2.1 1.9
1 1 3 125 2 1.7
1 2 1 72.9 1.6 1.1
1 2 1 38.1 1.6 1.1
1 2 1 8.3 1.3 1.8
1 2 1 31 1.6 0.8
1 2 1 36.3 1.3 1.2
1 2 2 16.8 1.9 1.8
1 2 2 4.8 2.2 1
1 2 2 20.4 1.4 1
1 2 2 31.8 1.9 0.4
1 2 2 5 1.6 1
1 2 3 17.9 2 2.2
1 2 3 14.9 2.6 2.5
1 2 3 5.3 1.6 0.5
1 2 3 17.1 2.2 1.3
1 2 3 17.1 1.9 1
1 3 1 67.3 1.6 1
1 3 1 11.3 1.4 0.7
1 3 1 4.2 1 1
1 3 1 9.7 1.1 0.8
1 3 1 20 1.4 0.8
1 3 2 4.3 1.4 1.3
1 3 2 4.1 1.6 2
1 3 2 11.4 1.5 0.3
1 3 2 12.5 1.4 0.2
1 3 2 11 1.5 0.9
1 3 3 59.7 1.9 2.4
1 3 3 81.4 2.3 2.4
1 3 3 32.2 1.9 1.6
1 3 3 24.9 1.8 1.2
1 3 3 50.9 2.1 1.2
2 1 1 185 1.6 1
2 1 1 146.2 1.7 1.2
2 1 1 239 1.8 1.4
2 1 1 141.9 2 2
2 1 1 303.7 1.9 1.8
2 1 2 53.6 1.9 0.6
2 1 2 424.8 2.1 1.5
2 1 2 30.3 1.9 0.4
2 1 2 291 2.1 1.5
2 1 2 471.6 2.1 2.1
2 1 3 132.8 1.9 1.4
2 1 3 124.5 2.1 1.7
2 1 3 484.4 2.1 2.1
2 1 3 251.3 2.2 2.2
2 1 3 598.6 2.1 2.2
2 2 1 15.6 1.4 0.9
2 2 1 107.6 1.7 1.1
2 2 1 172.2 1.8 1.6
2 2 1 210.2 1.9 1.9
2 2 1 699.7 2 2.2
2 2 2 20 1.7 0.2
2 2 2 86.2 2 0.9
2 2 2 36.8 2.1 0.3
2 2 2 434.4 2.1 2.3
2 2 2 209.5 2.2 1.5
2 2 3 37 2.1 1.6
2 2 3 94.8 2.1 1.8
2 2 3 220.2 2.2 2.1
2 2 3 17.7 2.3 0.8
2 2 3 145.3 2.1 1.9
2 3 1 33.4 1.5 0.8
2 3 1 87 1.7 1.2
2 3 1 98.9 1.8 1.8
2 3 1 176.6 2 2.2
2 3 1 180.7 1.9 1.9
2 3 2 20.8 1.8 0.3
2 3 2 226.2 2.1 1.1
2 3 2 227.8 2.1 1.7
2 3 2 40.8 2.1 0.4
2 3 2 83.4 2.1 0.9
2 3 3 35.7 2.1 1.2
2 3 3 191.9 2.1 1.8
2 3 3 203.2 2.2 2.2
2 3 3 143.3 2.2 2
2 3 3 70.7 2.1 1.7
3 1 1 190.7 2.1 2
3 1 1 456.5 2 2.2
3 1 1 520.9 2 2.2
3 1 1 535.5 2 2.1
3 1 1 894.3 2 2.3
3 1 2 265.2 2.1 1.7
3 1 2 392.2 2.1 2.3
3 1 2 241.3 2.1 1.4
3 1 2 156.9 2.1 1.7
3 1 2 277.2 2.1 1.7
3 1 3 431.6 2.1 2.1
3 1 3 669.9 2.2 2.3
3 1 3 342.5 2.1 2.1
3 1 3 572.3 2.1 2.2
3 1 3 435.2 2.1 2.1
3 2 1 498.6 2 2.1
3 2 1 1137.2 2.1 2.3
3 2 1 764.5 2.1 2.3
3 2 1 746.4 2.1 2.2
3 2 1 737.2 2.1 2.2
3 2 2 98.1 2 1.4
3 2 2 237.9 2.1 2.2
3 2 2 144 2 1.2
3 2 2 240.3 2.1 2.1
3 2 2 253.4 2.1 1.8
3 2 3 247.3 2.2 1.7
3 2 3 155.3 2.1 1.8
3 2 3 179.3 2.2 1.6
3 2 3 225.7 2.1 2.2
3 2 3 274.4 2.1 2.1
3 3 1 944.9 2.1 2.3
3 3 1 978.4 2.1 2.3
3 3 1 785.9 2 2.1
3 3 1 510.7 2 2.1
3 3 1 164.5 2 2.2
3 3 2 48.1 2 0.3
3 3 2 427.7 2.1 2.2
3 3 2 75 2 0.9
3 3 2 153 2.1 1.3
3 3 2 293.8 2.1 2.4
3 3 3 335.8 2.1 1.9
3 3 3 395.5 2.1 2.1
3 3 3 181.1 2.1 1.9
3 3 3 468.4 2.1 2.2
3 3 3 348 2.1 2.1
4 1 1 335.8 2 1.9
4 1 1 188.1 1.9 1.5
4 1 1 219 2 2
4 1 1 104.1 1.9 1.3
4 1 1 273.7 2 2
4 1 2 573.2 2.2 1.8
4 1 2 115.3 2.1 2.1
4 1 2 56.7 2 0.4
4 1 2 316.5 2.1 1.6
4 1 2 85.2 2.1 1.9
4 1 3 585 2.2 2.3
4 1 3 377.5 2.1 2
4 1 3 47.1 2.1 1.6
4 1 3 73.1 2 1.4
4 1 3 91.9 2 1.2
4 2 1 374.8 2 1.7
4 2 1 85.8 1.9 1.9
4 2 1 53.1 1.8 1.8
4 2 1 34.3 1.8 1.1
4 2 1 27 1.9 1.1
4 2 2 182.9 2.1 1.3
4 2 2 25.3 1.8 0.4
4 2 2 4.4 2.4 0.1
4 2 2 117.8 2.2 2.1
4 2 2 6.6 1.9 0.8
4 2 3 158.9 2.2 2.2
4 2 3 44.5 2 1
4 2 3 12.6 1.7 0.5
4 2 3 22.5 2.3 1
4 2 3 91 2 1.3
4 3 1 212.1 1.7 1.2
4 3 1 101.7 1.9 1.8
4 3 1 63.4 1.4 0.9
4 3 1 67.9 1.8 1.3
4 3 1 40.4 1.6 0.9
4 3 2 73 2.1 0.9
4 3 2 33.1 2.1 0.6
4 3 2 7.8 2 0.3
4 3 2 10.9 1.9 1.4
4 3 2 11.6 2 0.2
4 3 3 382.8 2.2 2.3
4 3 3 7 2.4 0.4
4 3 3 13.1 2 0.7
4 3 3 24.7 2.2 1.1
4 3 3 42.9 2 0.9
5 1 1 340 1.4 0.9
5 1 1 48.6 1.5 0.6
5 1 1 53.7 1.4 0.6
5 1 1 33.3 1.3 0.5
5 1 1 86 1.3 0.5
5 1 2 11.4 1.4 0.5
5 1 2 27 1.6 1
5 1 2 33.7 1.7 0.9
5 1 2 14.2 1.8 1.3
5 1 2 16.2 1.9 0.8
5 1 3 606 1.6 0.7
5 1 3 265.8 1.7 1
5 1 3 73.5 1.5 2.6
5 1 3 189 1.6 0.7
5 1 3 223.1 1.6 0.7
5 2 1 101.5 1.5 0.6
5 2 1 7.2 1.2 0.5
5 2 1 49.7 1.4 0.7
5 2 1 61.9 1.5 0.6
5 2 1 6.9 1.1 0.6
5 2 2 20.6 1.4 0.7
5 2 2 13.7 1.7 1
5 2 2 48.5 1.8 1
5 2 2 6.8 1.6 0.7
5 2 2 5.7 1.3 0.7
5 2 3 239.5 1.7 0.8
5 2 3 152.5 1.7 0.8
5 2 3 2065.4 1.5 1
5 2 3 112.7 1.6 0.6
5 2 3 104.5 1.6 0.7
5 3 1 40.7 1.3 0.5
5 3 1 82 1.4 0.9
5 3 1 20.6 1.3 1
5 3 1 21.4 1.4 0.6
5 3 1 29.6 1.4 0.6
5 3 2 4.3 1.1 0.5
5 3 2 38.5 1.3 1.3
5 3 2 22.9 1.7 1.1
5 3 2 10.3 1.7 1
5 3 2 4.9 1.3 0.6
5 3 3 216.8 1.7 0.8
5 3 3 220.7 1.7 0.9
5 3 3 51.8 1.3 1.3
5 3 3 161 1.6 0.7
5 3 3 144.4 1.6 0.6
6 1 1 79.9 2.1 1.8
6 1 1 295.3 2.1 2.1
6 1 1 136.4 2.1 2.1
6 1 1 177.3 2 2.1
6 1 1 116.4 1.7 1.1
6 1 2 93.1 2.1 2.2
6 1 2 385.9 2.1 2.3
6 1 2 318.6 2.1 1.8
6 1 2 20.3 2.1 1.4
6 1 2 131.5 2.1 1.4
6 1 3 53.9 2.2 2.7
6 1 3 156.9 2.2 1.8
6 1 3 344.7 2.1 2
6 1 3 15.5 2.6 0.6
6 1 3 33.7 2.1 1.3
6 2 1 137.8 1.8 1.2
6 2 1 104 2.1 2
6 2 1 151.7 2.1 2
6 2 1 15.4 2.1 1.6
6 2 1 31.5 1.9 2.1
6 2 2 8 2.1 1.5
6 2 2 23.9 2 0.9
6 2 2 14.8 2.2 0.4
6 2 2 8.6 1.8 0.4
6 2 2 18.2 2.2 0.3
6 2 3 35 2.2 2.7
6 2 3 61 2.2 1.7
6 2 3 136.9 2.1 2.3
6 2 3 3.1 8.5 0.2
6 2 3 8.7 1.6 1.1
6 3 1 155.7 2 2.2
6 3 1 48.5 1.7 0.9
6 3 1 77.7 2 1.5
6 3 1 45.6 1.6 1.2
6 3 1 142 1.6 1.1
6 3 2 5.6 1.4 1.3
6 3 2 7.4 2.2 0.1
6 3 2 66.4 1.9 1.4
6 3 2 6.8 2.3 0.7
6 3 2 10.3 1.9 2.2
6 3 3 42.9 2.1 2.5
6 3 3 48.3 2.3 1.2
6 3 3 62.6 2.3 1.4
6 3 3 6.9 2.2 0.3
6 3 3 3.9 1.8 0.5")
我是否忘记了要包含在 GLMM 中的任何内容？谢谢！
编辑 1：
我已将 Method 和 Kit 变量转换为因子格式：
Input_2$Metodo<-as.factor(Input_2$Metodo)
Input_2$Kit<-as.factor(Input_2$Kit)
获得不同的变暖：
Warning messages:
1: In checkConv(attr(opt,:
Model failed to converge with max|grad| = 0.00649539 (tol = 0.002,component 1)
2: In checkConv(attr(opt,:
Model is nearly unidentifiable: very large eigenvalue
- Rescale variables?
我该如何解决这些问题？
根据建议，我尝试重新调整我的价值观：
numcols <- grep("^c\\.",names(Input_2))
dfs <- Input_2
dfs[,numcols] <- scale(dfs[,numcols])
m1_sc <- update(glmer.1,data=dfs)
Obtaining the next warming message:
Warning messages:
1: In checkConv(attr(opt,:
Model failed to converge with max|grad| = 0.00649539 (tol = 0.002,component 1)
2: In checkConv(attr(opt,:
Model is nearly unidentifiable: very large eigenvalue
- Rescale variables?
Edit2：根据建议，我删除了我的交互 * for + 此外，我已将物种列转换为一个因子：
Input_2$Especie<-as.factor(Input_2$Especie)
glmer.1 <- glmer(Conc ~ Metodo + Kit +
(1|Especie),family = Gamma(link = "inverse"))
然后我得到下一次变暖：
Warning messages:
1: In checkConv(attr(opt,:
Model failed to converge with max|grad| = 0.00513611 (tol = 0.002,component 1)
2: In checkConv(attr(opt,:
Model is nearly unidentifiable: very large eigenvalue
- Rescale variables?
第二个建议是将 glmer 更改为 glm 以查看模型是否因更简单的策略而饱和（因为 glm 需要固定所有因子，我将物种变量更改为固定因子）：
glmer.1 <- glm(Conc ~ Metodo + Kit +Especie,family = Gamma(link = "inverse"))
当我执行它时，我没有问题。
如何使用 glmer 解决这些问题？我的错误是什么？
提前致谢
版权声明：本文内容由互联网用户自发贡献，该文观点与技术仅代表作者本人。本站仅提供信息存储空间服务，不拥有所有权，不承担相关法律责任。如发现本站有涉嫌侵权/违法违规的内容， 请发送邮件至 dio@foxmail.com 举报，一经查实，本站将立刻删除。