[BUG] Error plotting: Error in seq.default(x1, x2, by = ceiling((x2 - x1)/6)) : invalid '(to - from)/by' about drda

[GarrettJenkinson]

A related issue: if you try to plot two models built on two different sets of data (valuable for comparing curves across multiple conditions, etc) then the axes is wrong. Steps to reproduce:

fit <- drda(response ~ dose, data = voropm2,mean_function='loglogistic5',max_iter = 10000)
voropm2$dose <- voropm2$dose/2
fit2 <- drda(response ~ dose, data = voropm2,mean_function='loglogistic5',max_iter = 10000)
plot(fit,fit2,base="10")

The result has the x-axis incorrect with some values of dose appear to go negative (despite both datasets "stopping" at zero dose in their raw data).

[albertopessia]

Hi @GarrettJenkinson, thank you for your bug report. I believe I fixed both issues you reported.
Sorry it took so long, but I didn't receive any notification in my personal inbox and I only saw it today.

Updated code is now the devel branch. Install it with

remotes::install_github("albertopessia/drda@devel")

[GarrettJenkinson]

Thank you so much! Look forward to using the new version.

I have a question and perhaps feature request: is there reason to expect the reported IC50 confidence intervals are more conservative than the intervals drawn on these plots? It is not infrequent for me to see extremely wide confidence intervals reported from effective_dose and "reasonable" looking intervals by eye when doing a horizontal 'cut' on the plot. It might be nice to have a way to get the interval that agrees with what you see in the plot.

[albertopessia]

Hi @GarrettJenkinson, the confidence intervals you talk about are for two different quantities and they are expected to be different.

The confidence interval for the whole curve, as shown in the plot, refers to the y-axis (think of it as many vertical segments). Horizontal slices are not to be looked at as they are not meaningful.

Confidence intervals reported by effective_dose are indeed on the x-axis but unrelated from horizontal slices of the plot.

Run this code to understand what's going on:

library(drda)
fit <- drda(response ~ log_dose, data = voropm2, mean_function = "l5")
plot(fit, midpoint = FALSE)
abline(h = 0.02, lty = 2)
effective_dose(fit, y = 0.02, type = "absolute", level = 0.99)
abline(v = c(6.918, 7.707), lty = 2)

If you look at the horizontal slice the upper bound is infinite. However, by looking at the data, we understand that the curve will reach the value 0.02 almost certainly between 6.92 and 7.71.

Hope this helps.