lme4模型显示交互显著，effects包生成的效应图却无交互？

我用R语言的lme4包给数据集拟合了广义线性混合模型，模型输出显示period与Adjunct存在显著交互作用（p<0.001）。之后我用effects包提取二者的效应，再用ggplot绘制交互效应图，但生成的图里看不到明显的交互趋势。怀疑是脚本有问题，相关数据集已备好。

提取效应脚本 #

var.adSem <- data.frame(Effect(c("period", "AdjunctSem"), m7))
head(var.adSem)
var.adSem$period <- droplevels(var.adSem$period)
levels(var.adSem$period)
var.adSem$AdjunctSem <- droplevels(var.adSem$AdjunctSem)
levels(var.adSem$AdjunctSem)

绘图脚本 #

perc <- function(x) {paste(100 * x, "%")}
Plot1<-(ggplot(var.adSem,aes(x=factor(period, level=c('period_1', 'period_2','period_3')),fit, shape=AdjunctSem))
        + geom_hline(yintercept=.5, linetype="dashed", alpha=.5)  
        + geom_errorbar(aes(ymin=fit-se,ymax=fit+se),width=.2,position = position_dodge(0.3))
        #    + annotate("rect", xmin = c(4.5, 6.5), xmax = c(5.5, 9.5),
        #           ymin = -Inf, ymax = Inf,
        #           alpha = 0.2, fill = c("#6ac2ee"))
        + geom_point(size=4, position = position_dodge(0.3))
        + labs(y="Probability of the ba-construction", x="period")
        + theme(#plot.title = element_text(size=20, face="bold", vjust=2),
          plot.title   = element_blank(),
          axis.title.x = element_text(size=14, vjust=-0.5),
          axis.text.x  = element_text(angle=90, hjust=1, vjust=0.5),
          axis.title.y = element_text(size=14, vjust=1.5),
          legend.position = "top")
        + scale_y_continuous(label=function(x){paste(100*x,"%")}, 
                             breaks=seq(0,1,.1),
                             limits=c(0,1)))

可能的错误点及修正方案 #

• 变量名不匹配：模型交互项是period和Adjunct，但提取效应时用的是AdjunctSem，需核对模型m7的公式，确保提取的变量与模型中的交互项一致。
• y轴映射缺失：ggplot的aes中未明确指定y=fit，这会导致绘图逻辑错误，应修正为aes(x=..., y=fit, shape=AdjunctSem)。
• 缺少趋势折线：仅用散点和误差棒无法直观展示交互趋势，需添加geom_line()并设置group=AdjunctSem和position=position_dodge(0.3)，对齐散点与折线。
• 因子水平重复处理：提取效应后已用droplevels()处理水平，绘图时无需重复用factor()指定，可改用scale_x_discrete()统一设置水平顺序。

修正后的绘图脚本示例 #

perc <- function(x) {paste(100 * x, "%")}
Plot1 <- ggplot(var.adSem, aes(x = period, y = fit, shape = AdjunctSem, group = AdjunctSem)) +
  geom_hline(yintercept = .5, linetype = "dashed", alpha = .5) +
  geom_errorbar(aes(ymin = fit - se, ymax = fit + se), width = .2, position = position_dodge(0.3)) +
  geom_point(size = 4, position = position_dodge(0.3)) +
  geom_line(position = position_dodge(0.3)) +  # 添加折线展示交互趋势
  labs(y = "Probability of the ba-construction", x = "period") +
  theme(plot.title = element_blank(),
        axis.title.x = element_text(size = 14, vjust = -0.5),
        axis.text.x = element_text(angle = 90, hjust = 1, vjust = 0.5),
        axis.title.y = element_text(size = 14, vjust = 1.5),
        legend.position = "top") +
  scale_y_continuous(label = perc, breaks = seq(0, 1, .1), limits = c(0, 1)) +
  scale_x_discrete(levels = c('period_1', 'period_2','period_3'))  # 统一设置x轴水平顺序

内容的提问来源于stack exchange，提问作者nick.yili9393