customizing spaghetti plot-CodePudding

I am trying to create a custom spaghetti plot

Assuming this is my dataset

data(package = "geepack")
data("dietox") 
head(dietox)




    head(dietox, n =20)
    Pig    Evit    Cu Litter    Start   Weight       Feed Time
   4601 Evit000 Cu000      1 26.50000 26.50000         NA    1
   4601 Evit000 Cu000      1 26.50000 27.59999   5.200005    2
   4601 Evit000 Cu000      1 26.50000 36.50000  17.600000    3
   4601 Evit000 Cu000      1 26.50000 40.29999  28.500000    4
   4601 Evit000 Cu000      1 26.50000 49.09998  45.200001    5
   4601 Evit000 Cu000      1 26.50000 55.39999  56.900002    6
   4601 Evit000 Cu000      1 26.50000 59.59998  71.700005    7
   4601 Evit000 Cu000      1 26.50000 67.00000  86.800001    8
   4601 Evit000 Cu000      1 26.50000 76.59998 104.900002    9
   4601 Evit000 Cu000      1 26.50000 86.50000 123.000000   10
   4601 Evit000 Cu000      1 26.50000 91.59998 140.900002   11
   4601 Evit000 Cu000      1 26.50000 98.59998 160.000000   12
   4602 Evit000 Cu035      1 28.29999 28.29999         NA    1
   4602 Evit000 Cu035      1 28.29999 30.09999   3.300003    2
   4602 Evit000 Cu035      1 28.29999 38.29999  13.200001    3
   4602 Evit000 Cu035      1 28.29999 44.50000  26.100000    4
   4602 Evit000 Cu035      1 28.29999 51.59998  43.600002    5
   4602 Evit000 Cu035      1 28.29999 57.59998  55.200001    6


 ggplot(data = dietox, aes(x = Time, y = Weight, group = Pig))  
 geom_line()

This creates a plot like this

What I am interested in , is creating a gee predicted regression lines at two time points, 1)0- 6, 2) 6-12. The expected plot should look like this.

mf <- formula(Weight ~ Cu * (Time   I(Time^2)   I(Time^3)))
gee1 <- geeglm(mf, data=dietox, id=Pig, family="gaussian", corstr="ar1")
gee1

I would really appreciate your help.

CodePudding user response：

Your model takes the variable Cu into account, so each of your periods would have 3 different lines - one for each level of Cu. Since this is just sample data taken from the package, and you are looking for a single line in each period, I assume that your actual data is a function of a single variable.

The way to plot this is to manually predict lines for the two separate periods and plot them:

library(geepack)
library(ggplot2)
library(hrbrthemes)

data("dietox") 

mf   <- formula(Weight ~ Time   I(Time^2)   I(Time^3))

gee1 <- geeglm(mf, data = dietox[dietox$Time <= 6,], 
               id = Pig, family = "gaussian", corstr ="ar1")
gee2 <- geeglm(mf, data = dietox[dietox$Time >= 6,], 
               id = Pig, family = "gaussian", corstr ="ar1")
dietox$predictions <- numeric(1)
dietox$period <- "A"
dietox$predictions[dietox$Time <= 6] <- predict(gee1)
dietox$predictions[dietox$Time >= 6] <- predict(gee2)
dietox$period[dietox$Time >= 6] <- "B"


ggplot(data = dietox, aes(x = Time, y = Weight, group = Pig))  
 geom_line(aes(group = Pig), size = 2, alpha = 0.2)  
 geom_line(aes(y = predictions, group = interaction(Pig, period),
               colour = period), size = 2)  
 theme_ipsum()

^{Created on 2022-03-09 by the reprex package (v2.0.1)}