将拟合回归线添加到 R plotly 小提琴

Question

I have xy data for three group s across three integer age s:

set.seed(1)
df <- data.frame(value = c(rnorm(500,8,1),rnorm(600,6,1.5),rnorm(400,4,0.5),rnorm(500,2,2),rnorm(400,4,1),rnorm(600,7,0.5),rnorm(500,3,1),rnorm(500,3,1),rnorm(500,3,1)),
                 age = c(rep(3,500),rep(8,600),rep(24,400),rep(3,500),rep(8,400),rep(24,600),rep(3,500),rep(8,500),rep(24,500)),
                 group = c(rep("A",1500),rep("B",1500),rep("C",1500)))

My purpose is to use R 's plotly to plot the value s as violin s, like this:

library(plotly)
library(dplyr)
df$age <- factor(df$age)
plot_ly(x=df$group,y=df$value,type='violin',name=df$age,color=df$age,box=list(visible=T)) %>%
  layout(violinmode='group')

And add to each group the lm fitted line, and if possible also the standard error lines.

So I first add to df the fitted lines per each group :

df$age <- as.integer(df$age)
df$fitted.value <- unlist(lapply(c("A","B","C"),function(g) lm(value ~ age,dplyr::filter(df,group == g)) %>% fitted.values()))

Then I tried to only do a single group 's violin adding it's fitted line using:

df <- df %>% dplyr::filter(group == "A")
plot_ly(x=df$age,y=df$value,type='violin',color=df$age,box=list(visible=T)) %>%
  add_trace(x=df$age,y=df$fitted.value,mode="lines")

But this gives this long list of warnings :

Warning messages:
1: 'violin' objects don't have these attributes: 'mode'
Valid attributes include:
'type', 'visible', 'showlegend', 'legendgroup', 'opacity', 'uid', 'ids', 'customdata', 'meta', 'selectedpoints', 'hoverinfo', 'hoverlabel', 'stream', 'transforms', 'uirevision', 'y', 'x', 'x0', 'y0', 'name', 'orientation', 'bandwidth', 'scalegroup', 'scalemode', 'spanmode', 'span', 'line', 'fillcolor', 'points', 'jitter', 'pointpos', 'width', 'marker', 'text', 'hovertext', 'hovertemplate', 'box', 'meanline', 'side', 'offsetgroup', 'alignmentgroup', 'selected', 'unselected', 'hoveron', 'xaxis', 'yaxis', 'idssrc', 'customdatasrc', 'metasrc', 'hoverinfosrc', 'ysrc', 'xsrc', 'textsrc', 'hovertextsrc', 'hovertemplatesrc', 'key', 'set', 'frame', 'transforms', '_isNestedKey', '_isSimpleKey', '_isGraticule', '_bbox'
 
2: 'violin' objects don't have these attributes: 'mode'
Valid attributes include:
'type', 'visible', 'showlegend', 'legendgroup', 'opacity', 'uid', 'ids', 'customdata', 'meta', 'selectedpoints', 'hoverinfo', 'hoverlabel', 'stream', 'transforms', 'uirevision', 'y', 'x', 'x0', 'y0', 'name', 'orientation', 'bandwidth', 'scalegroup', 'scalemode', 'spanmode', 'span', 'line', 'fillcolor', 'points', 'jitter', 'pointpos', 'width', 'marker', 'text', 'hovertext', 'hovertemplate', 'box', 'meanline', 'side', 'offsetgroup', 'alignmentgroup', 'selected', 'unselected', 'hoveron', 'xaxis', 'yaxis', 'idssrc', 'customdatasrc', 'metasrc', 'hoverinfosrc', 'ysrc', 'xsrc', 'textsrc', 'hovertextsrc', 'hovertemplatesrc', 'key', 'set', 'frame', 'transforms', '_isNestedKey', '_isSimpleKey', '_isGraticule', '_bbox'
 
3: 'violin' objects don't have these attributes: 'mode'
Valid attributes include:
'type', 'visible', 'showlegend', 'legendgroup', 'opacity', 'uid', 'ids', 'customdata', 'meta', 'selectedpoints', 'hoverinfo', 'hoverlabel', 'stream', 'transforms', 'uirevision', 'y', 'x', 'x0', 'y0', 'name', 'orientation', 'bandwidth', 'scalegroup', 'scalemode', 'spanmode', 'span', 'line', 'fillcolor', 'points', 'jitter', 'pointpos', 'width', 'marker', 'text', 'hovertext', 'hovertemplate', 'box', 'meanline', 'side', 'offsetgroup', 'alignmentgroup', 'selected', 'unselected', 'hoveron', 'xaxis', 'yaxis', 'idssrc', 'customdatasrc', 'metasrc', 'hoverinfosrc', 'ysrc', 'xsrc', 'textsrc', 'hovertextsrc', 'hovertemplatesrc', 'key', 'set', 'frame', 'transforms', '_isNestedKey', '_isSimpleKey', '_isGraticule', '_bbox'

And this undesired plot:

Any idea how to add the trend lines for all group s and optimally also their standard error lines?

Answer 1

Maybe you can use this solution. It starts by making a new df with an additional variable describing the group and the subgroup .

Another DF builds the mean values of group_by(group, moreA) .

The plot is made with ggplot and the geom_violin is filled with data from the first DF and geom_point from the second DF sum_res as well as the geom_smooth does. This geom_smooth takes the mean values from sum_res to make the fit.

After this the ggplot object is brought into ggplotly .

Unfortunately the ggplotly behaviour like hovering does not come out in this reprex. But it is working in RStudio

library(plotly)
library(tidyverse)


ddf <- df %>% #glimpse()
  mutate(moreA = case_when(
    age == 3 ~ paste0(group,'3'),
    age == 8 ~ paste0(group,'8'),
    age == 24 ~ paste0(group,'24'),
  )) 

# this will bring the right order
ddf$moreA <- factor(ddf$moreA, levels = unique(ddf$moreA))

sum_res <- ddf %>% #
  group_by(group, moreA) %>% 
  summarise(meanA = mean(value))
    
p <- ggplot(ddf, aes(x = moreA, y = value, color = moreA)) +
  geom_violin() +
  geom_point(data = sum_res, mapping = aes(x = moreA, y = meanA), size = 1) +
  geom_smooth(data = sum_res, 
              mapping = aes(x = moreA, y = meanA, group = group, color = group),
              method='lm', size = 1, se =F) +
  theme_bw() +
  theme(legend.position = 'none')
#p

ggplotly(p)