1. 技术文章
  2. Reproducing drc::plot.drc with ggplot2

Reproducing drc::plot.drc with ggplot2

由 我与影子孤独终老i 提交于 2019-12-05 15:52:40

As noted in the question, there is an issue related to drm shuffling the order of factor levels. Un-shuffling this mess proved more tricky than I expected.

In the end I approached this by calling the drm function once per factor level to build up a table of results one factor level at a time.

Doing it this long-winded way uncovered the fact that your 1st plot from plot.drc and the ggplot version are both incorrect.

Let's start by wrapping your function call to drm() inside another wrapper function, to facilitate calling it repeatedly for each trace:

drcmod <- function(dt1){
  drm(formula   = Germinated ~ Start + End
    , curveid   = TempV
    , data      = dt1
    , fct       = LL.2()
    , type      = "event"
    , bcVal     = NULL
    , bcAdd     = 0
    , na.action = na.fail
    , robust    = "mean"
    , logDose   = NULL
    , control   = drmc(
      constr      = FALSE
      , errorm      = TRUE
      , maxIt       = 1500
      , method      = "BFGS"
      , noMessage   = FALSE
      , relTol      = 1e-07
      , rmNA        = FALSE
      , useD        = FALSE
      , trace       = FALSE
      , otrace      = FALSE
      , warnVal     = -1
      , dscaleThres = 1e-15
      , rscaleThres = 1e-15
    )
    , lowerl    = NULL
    , upperl    = NULL
    , separate  = FALSE
    , pshifts   = NULL 
  )
}

Now we can use this wrapper to fit the drc model to each factor level in turn:

dt2 <- data.table()
for (i in 1:nlevels(dt1$TempV)) {
  dt <- dt1[TempV==levels(TempV)[i]]
  dt[, TempV:=as.character(TempV)]
  dt[, Germ_frac := mean(Germinated)/450, by=.(Start)]
  dt[, cum_Germinated := cumsum(Germ_frac)]
  dt[, Pred := c(predict(object=drcmod(dt)), NA)] 
  dt2 <- rbind(dt2, dt)
}

and plot:

ggplot(dt2[End != Inf], aes(x=End, y=cum_Germinated, group=TempV, color=TempV, shape=TempV)) + 
  geom_point() +
  geom_line(aes(y = Pred)) +
  scale_shape_manual(values=seq(0, 13)) +
  labs(x = "Time (Hours)", y = "Proportion Germinated", shape="Temp", color="Temp") +
  theme_bw()

Edit

If we run the original code in the question using a subset of the data with fewer factor levels, for example using

dt1 <- dt1[TempV %in% levels(TempV)[1:5],]
dt1 <- droplevels(dt1)

all the plots (the 2 versions in OP, and the version in this answer) give the same result. The discrepancies only seem to arise when a large number of factor levels are used. The fact that both the ggplot and the plot.drc in OP give incorrect matching of traces to factor levels indicates that the problem is most likely to be in the drm() function, rather than in plot.drc.

标签
r
ggplot2
logistic-regression
drc
易学教程内所有资源均来自网络或用户发布的内容，如有违反法律规定的内容欢迎反馈！
该文章没有解决你所遇到的问题?点击提问,说说你的问题,让更多的人一起探讨吧!