Non-linear regression line and Computation failed in `stat_smooth()`: argument "p" is missing, with no default

[remi133]

Good Evening (or morning) everyone,

I have been trying to fit a non-linear regression line into my standard curve. However, I am getting the following error:

I have been using the following code:

 stat_smooth(method= "nlm", 
                formula = y ~ a*x^b,
                method.args = list(start = c(a = 470, b = 460)),
                se=FALSE)+

I am not sure whether the value 'a' and 'b' are correct, as I have taken them from the regline equation code:

stat_cor(label.y = c(825),
              label.x = c(0.88), 
          aes(label = paste(..rr.label.., ..p.label.., sep = "~`,`~")))+
   stat_regline_equation(label.x=0.88, label.y=750)+

I have never been fitting a non-linear regression line. However, it is crucial in this experiment as I have got two variables only (concentration and the ratio) in a non-linear relationship.

This is the following code used

library(tidyverse)
library(tidyr)
library(dplyr)
library(readr)
library(ggplot2)
library(ggpubr)
library(gridExtra)
library(cowplot)
ggplot(data = STD, aes(x = Absorbance, y = STD)) +
  labs(title = "Quantifying PGD2 in cell culture lysates and its enzymatic reactions ",
       caption = "PGD2 ELISA")+
    geom_point(colour = "#69b3a2")+
    stat_smooth(method= "nlm", 
                formula = y ~ a*x^b,
                method.args = list(start = c(a = 470, b = 460)),
                se=FALSE)+
    xlab(expression(paste("%B/"~B[0])))+
    ylab(expression(paste("Prostaglandin"~ D[2], ~~ " MOX Concentration (pg/ml) ")))+
    
   theme(plot.background =  element_rect(fill = "transparent"),
         panel.background = element_blank(),
         panel.grid.major = element_blank(),
         panel.grid.minor = element_blank(),
         axis.line = element_line(colour = "black"))+
  
   theme(legend.spacing.y = unit(0.01, "cm"))+
   theme(legend.position = c(0.77, .91),
         legend.background = element_rect(colour = NA, fill = NA))+
   theme(plot.title = element_text(size = 12, face = "bold.italic"),
         plot.caption = element_text(hjust = 0))

That produced the following outcome

Before that, I have added a line, but it is not a non-linear regression line tho.

Moreover, I have been thinking about whether the non-linear equation fits my graph, but it fulfills the legibility of the equation, so this is not the case.

DataUsed

Thank you so much for your help.

[startz]

I believe (but I'm not sure) that for nlm() the starting values are called p rather than start.

Also, are you sure that you want to raise x to the power 460? That's likely to give you either a very small number of a very large number.

[remi133]

Thank you for your answer, unfortunately, that is not the case. After changing the starting value from start to p the error changes and says

Warning: Computation failed in `stat_smooth()`:
could not find function "f"

Answering your second question, I am not sure. Therefore, I am trying to find the right equation for my graph. I think I will use the Michaelis-Menten equation.

[startz]

Sorry that I wasn't more helpful on the main question. On the secondary issue, note that

> 1.1^460
[1] 1.098083e+19
> 0.9^460
[1] 8.944464e-22

[remi133]

No worries. The data you are showing to me what should it tell me?

I can just update my question. The equation above that I was certain that is correct is wrong, particularly I must use the four parametric equation: y=d + a-d/1+ (x/c)^2, however after updating my code chunk I am getting the following warning: Warning: Computation failed in stat_smooth(): parameters without starting value in data: 'd'. Any ideas on how to overcome such warning?

[startz]

In the equation you had, the starting parameters were suggesting very strange values for y. Since you've changed the equation, that probably doesn't matter.

The problem is that stat_smooth() isn't finding the starting values for your nonlinear estimate. Check that d, a, and c are all being fed in properly.

And as a wild guess (so quite likely wrong), are you sure you want to be using nlm rather than nls?

[remi133]

I was surprised that I was getting such strange values for y. Nevertheless, it does not matter if I use nls , or nlm in both examples I am getting this same warning, unfortunately. And yes, I have no idea why is that stat_smooth() is not finding the starting values. Probably I could have used Log logistic functions, however, after typing summary(mydata) it does not find b, c, d and e intercepts

[startz]

As a debugging suggestion, you might try running your nonlinear estimate without embedding it in stat_smooth. That will help identify whether the problem is in feeding instructions to stat_smooth or in the underlying estimate.

[remi133]

Ok, so I have found a solution. I installed the drc package in which the four parametric function is included. I set up my data model <- drm(STD ~ Absorbance, fct = LL.4(), data = STD), then plot(model) , and I got

I know it requires some alternations to make it look more professional, but it is just a cosmetic thing that I should be fine to do. Thank you so much for your time @startz