这是从我的完整数据框中随机选择的数据：
canopy<-structure(list(Stage = structure(c(6L, 5L, 3L, 6L, 7L, 5L, 4L,
7L, 2L, 7L, 5L, 1L, 1L, 4L, 3L, 6L, 5L, 7L, 4L, 4L), .Label = c("milpa",
"robir", "jurup che", "pak che kor", "mehen che", "nu kux che",
"tam che"), class = c("ordered", "factor")), ID = c(44L, 34L,
18L, 64L, 54L, 59L, 28L, 51L, 11L, 56L, 33L, 1L, 7L, 25L, 58L,
48L, 36L, 51L, 27L, 66L), Sample = c(4L, 2L, 2L, 10L, 6L, 9L,
4L, 3L, 3L, 8L, 1L, 1L, 7L, 1L, 10L, 8L, 4L, 3L, 3L, 10L), Subsample = c(2L,
3L, 4L, 3L, 2L, 1L, 3L, 2L, 4L, 3L, 1L, 3L, 2L, 4L, 1L, 1L, 3L,
1L, 1L, 4L), Size..ha. = c(0.5, 0.5, 0.5, 0.5, 6, 0.5, 0.5, 0.25,
0.5, 6, 1, 1, 0.5, 2, 1, 0.5, 1, 0.25, 0.5, 2), Avg.Subsample.Canopy = c(94.8,
94.8, 97.92, 96.88, 97.14, 92.46, 93.24, 97.4, 25.64, 97.4, 94.8,
33.7, 13.42, 98.18, 85.44, 96.36, 97.4, 95.58, 85.7, 92.2), dec = c(0.948,
0.948, 0.9792, 0.9688, 0.9714, 0.9246, 0.9324, 0.974, 0.2564,
0.974, 0.948, 0.337, 0.1342, 0.9818, 0.8544, 0.9636, 0.974, 0.9558,
0.857, 0.922)), .Names = c("Stage", "ID", "Sample", "Subsample",
"Size..ha.", "Avg.Subsample.Canopy", "dec"), row.names = c(693L,
537L, 285L, 1017L, 853L, 929L, 441L, 805L, 173L, 889L, 513L,
9L, 101L, 397L, 913L, 753L, 569L, 801L, 417L, 1053L), class = "data.frame")
我正在尝试将dec的GLMM编码为Stage和Size的函数。 GLMM是必要的，因为每行代表在较大的样本区域内测量的子样本。我也使用二项分布给出dec是比例数据。
我试过这个模型：
canopy.binomial.mod<-glmer(dec~Stage*Size..ha.+(1|Sample),family="binomial",data=canopy)
summary(canopy.binomial.mod)
但得到错误：
pwrssUpdate中的错误（pp，resp，tol = tolPwrss，GQmat = GQmat，compDev = compDev，:( maxstephalfit）PIRLS步长减半未能减少pwrssUpdate中的偏差
我在网上看到这可能是需要扩展预测变量的结果，所以我试过了：
cs. <- function(x) scale(x,scale=TRUE,center=TRUE)
canopy.binomial.mod<-glmer(dec~Stage*cs.(Size..ha.)+(1|Sample),family="binomial",data=canopy.rmna)
summary(canopy.binomial.mod)
这似乎没什么帮助。我也认为，由于预测变量太多，我可能会过多地询问模型并且它没有收敛，所以让我们删除Size变量，这对我来说不太重要。
canopy.binomial.mod<-glmer(dec~Stage+(1|Sample),family="binomial",data=canopy.rmna)
summary(canopy.binomial.mod)
仍然没有运气。任何想法如何解决这个问题？