3

I am working on an exercise question, and I am a bit stuck unsure on how to proceed. I was given the data below and asked to graphically display the relationship between dose and kill rate.

enter image description here

Using R, I created a data frame to represent the values from the table with the following commands,

dose <- c("1", "2", "3", "4", "5", "6")
insecticide <- c("1", "2", "3")
data.df <- expand.grid(dose=dose, insecticides=insecticide)

kills <- c(3, 5, 19, 19, 24, 35,
           2, 14, 20, 27, 41, 40,
           28, 37, 46, 48, 48, 50)


total_beetles <- c(50, 49, 47, 38, 49, 50,
                   50, 49, 50, 50, 50, 50,
                   50, 50, 50, 50, 50, 50)


beetles.df <- as.data.frame(data.df) %>%
  mutate(kills = kills, beetles=total_beetles, 
         killrate=kills/beetles)

I then displayed the relationship between dose and kill rate graphically using a facet wrap to view the insecticides as well.

ggplot(data=beetles.df, mapping=aes(x=dose, y = killrate, 
       color=insecticides)) + geom_point() + 
       facet_wrap(vars(insecticides))

Now for this exercise question, I was asked to, plot the linear logistic fitted curve for each of the insecticides plus the combination, but I am not sure I can plot a logistic curve in this case since wouldn't my response variable have to be binary to apply a logistic curve. I am just a bit confused maybe on the wording of this question.

kjetil b halvorsen
  • 77,844
Harry Lofi
  • 303

1 Answers1

1

but I am not sure I can plot a logistic curve in this case since wouldn't my response variable have to be binary to apply a logistic curve.

You have multiple observations at the same value of insecticide. So you get counts per total number of observations. Such counts are binomial distributed which is a sum of Bernoulli distributed variables. The count variable is actually under the hood a binary variable (a sum of several binary variables).

See also: In using the cbind() function in R for a logistic regression on a $2 \times 2$ table, what is the explicit functional form of the regression equation?

Sextus Empiricus
  • 77,915