如何绘制混合模型的结果

Question

我在R中使用lme4来适应混合模型

lmer(value~status+(1|experiment)))

值是连续的,状态(N/D/R)和实验是因素,我得到

Linear mixed model fit by REML 
Formula: value ~ status + (1 | experiment) 
  AIC   BIC logLik deviance REMLdev
 29.1 46.98 -9.548    5.911    19.1
Random effects:
 Groups     Name        Variance Std.Dev.
 experiment (Intercept) 0.065526 0.25598 
 Residual               0.053029 0.23028 
Number of obs: 264, groups: experiment, 10

Fixed effects:
            Estimate Std. Error t value
(Intercept)  2.78004    0.08448   32.91
statusD      0.20493    0.03389    6.05
statusR      0.88690    0.03583   24.76

Correlation of Fixed Effects:
        (Intr) statsD
statusD -0.204       
statusR -0.193  0.476

我想以图形方式表示固定效应评估.但是这些对象似乎没有绘图功能.有什么方法可以用图形方式描绘固定效果吗？

Answer 1

使用coefplot2(在r-forge上):

从@Thierry窃取模拟代码:

set.seed(101)
dataset <- expand.grid(experiment = factor(seq_len(10)), 
    status = factor(c("N", "D", "R"), levels = c("N", "D", "R")), 
    reps = seq_len(10))
X <- model.matrix(~status,dataset)
dataset <- transform(dataset, 
    value=rnorm(nrow(dataset), sd = 0.23) +   ## residual
    rnorm(length(levels(experiment)), sd = 0.256)[experiment] +  ## block effects
    X %*% c(2.78,0.205,0.887))  ## fixed effects

适合型号:

library(lme4)
model <- lmer(value~status+(1|experiment), data = dataset)

情节:

install.packages("coefplot2",repos="http://r-forge.r-project.org")
library(coefplot2)
coefplot2(model)

编辑:

我经常遇到R-Forge构建问题.如果R-Forge构建不起作用,则此回退应该有效:

install.packages("coefplot2",
  repos="http://www.math.mcmaster.ca/bolker/R",
  type="source")

请注意,coda必须已安装依赖项.

Answer 2

我喜欢系数置信区间图,但考虑一些额外的图来理解固定效应可能是有用的.

从@Thierry窃取模拟代码:

library(ggplot2)
library(lme4)
library(multcomp)
dataset <- expand.grid(experiment = factor(seq_len(10)), status = factor(c("N", "D", "R"), levels = c("N", "D", "R")), reps = seq_len(10))
dataset$value <- rnorm(nrow(dataset), sd = 0.23) + with(dataset, rnorm(length(levels(experiment)), sd = 0.256)[experiment] + ifelse(status == "D", 0.205, ifelse(status == "R", 0.887, 0))) + 2.78
model <- lmer(value~status+(1|experiment), data = dataset)

看看数据的结构......看起来平衡..

library(plotrix); sizetree(dataset[,c(1,2)])

跟踪固定效应之间的相关性可能很有趣,特别是如果您适合不同的相关结构.以下链接提供了一些很酷的代码......

http://hlplab.wordpress.com/2012/03/20/correlation-plot-matrices-using-the-ellipse-library/

my.plotcorr(
matrix(c(1,     .891,   .891,
        .891,   1,      .891,
        .891,   .891,   1), nrow=3)
)

最后,看看10个实验的变异性以及实验中"状态"的变异性似乎是相关的.我正在研究这个代码,因为我在不平衡数据上打破它,但这个想法是......

My2Boxes(m=4,f1=dataset$experiment,f2=dataset$status,x=dataset$value,color=c("red","yellow","green"))

最后,已经提到过的Piniero和Bates(2000)的书强烈赞成格子从我撇去的那些小东西.所以你可能会给它一个机会.也许像绘制原始数据......

lattice::xyplot(value~status | experiment, groups=experiment, data=dataset, type=c('p','r'), auto.key=F)

然后绘制拟合值......

lattice::xyplot(fitted(model)~status | experiment, groups=experiment, data=dataset, type=c('p','r'), auto.key=F)

Answer 3

以下是一些建议.

library(ggplot2)
library(lme4)
library(multcomp)
# Creating datasets to get same results as question
dataset <- expand.grid(experiment = factor(seq_len(10)),
                       status = factor(c("N", "D", "R"),
                       levels = c("N", "D", "R")),
                       reps = seq_len(10))
dataset$value <- rnorm(nrow(dataset), sd = 0.23) + 
                   with(dataset, rnorm(length(levels(experiment)),
                        sd = 0.256)[experiment] +
                   ifelse(status == "D", 0.205,
                          ifelse(status == "R", 0.887, 0))) +
                   2.78

# Fitting model
model <- lmer(value~status+(1|experiment), data = dataset)

# First possibility
tmp <- as.data.frame(confint(glht(model, mcp(status = "Tukey")))$confint)
tmp$Comparison <- rownames(tmp)
ggplot(tmp, aes(x = Comparison, y = Estimate, ymin = lwr, ymax = upr)) +
  geom_errorbar() + geom_point()

# Second possibility
tmp <- as.data.frame(confint(glht(model))$confint)
tmp$Comparison <- rownames(tmp)
ggplot(tmp, aes(x = Comparison, y = Estimate, ymin = lwr, ymax = upr)) +
  geom_errorbar() + geom_point()

# Third possibility
model <- lmer(value ~ 0 + status + (1|experiment), data = dataset)
tmp <- as.data.frame(confint(glht(model))$confint)
tmp$Comparison <- rownames(tmp)
ggplot(tmp, aes(x = Comparison, y = Estimate, ymin = lwr, ymax = upr)) +
  geom_errorbar() + geom_point()