r - Plotting Piecewise growth curves

Question

I am trying to plot a piecewise growth curve similar to this first plot. I used the separate slopes coding scheme and placed a breakpoint at time 2

| time | 0 | 1 | 2 | 5 | 10 | 15 | 20|

| time1 | 0 | 1 | 2 | 2 | 2 | 2 | 2 |

| time2 | 0 | 0 | 0 | 1 | 2 | 3 | 4 |

I used the following code to create my growth model

m1 <- lmer(sdmtwr ~ time1 + time2 + (time1 | id) + (0 + time2 | id), data = SDMT, REML = FALSE)

I'm also exploring an interaction with a 2-level categorical predictor with the following code

m2 <- lmer(sdmtwr ~ (time1 + time2)*edu + (time1 | id) + (0 + time2 | id), data = SDMT, REML = FALSE)

I've attempted to create the plots with the ggplot2, sjPlot, and effects packages to no avail, and I am at a loss due to limited programming experience. I have only ever been able to plot segments separately for both the baseline and interaction models.

If anyone could provide assistance on the appropriate code, I would appreciate it!

Edit: Here is the dput summary (edited for length to show edu, time1, and time2)

> dput(sdmt)
structure(list(id = c(3L, 3L, 3L, 3L, 3L, 4L, 4L, 4L, 5L, 6L, 
6L, 6L, 28L, 28L, 28L, 28L, 28L, 28L, 28L, 62L, 62L, 62L, 62L, 
108L, 108L, 108L, 108L, 119L, 119L, 120L, 120L, 120L, 120L, 132L, 
132L, 132L, 132L, 132L, 148L, 148L, 148L, 148L, 148L, 148L, 175L, 
175L, 175L, 178L, 178L, 178L, 178L, 201L, 201L, 201L, 201L, 201L, 
201L, 201L, 253L, 253L, 253L, 253L, 327L, 327L, 327L, 327L, 336L, 
336L, 336L, 336L, 336L, 336L, 343L, 343L, 360L, 360L, 360L, 366L, 
366L, 366L), time = c(0L, 2L, 10L, 15L, 20L, 5L, 10L, 15L, 2L, 
2L, 15L, 20L, 0L, 1L, 2L, 5L, 10L, 15L, 20L, 5L, 10L, 15L, 20L, 
0L, 2L, 15L, 20L, 0L, 2L, 0L, 10L, 15L, 20L, 0L, 1L, 5L, 10L, 
20L, 1L, 2L, 5L, 10L, 15L, 20L, 0L, 1L, 2L, 0L, 1L, 2L, 5L, 0L, 
1L, 2L, 5L, 10L, 15L, 20L, 0L, 1L, 5L, 15L, 0L, 1L, 10L, 20L, 
0L, 1L, 5L, 10L, 15L, 20L, 0L, 10L, 1L, 5L, 10L, 0L, 10L, 15L
), sdmtwr = c(20L, 24L, 18L, 19L, 9L, 17L, 24L, 17L, 41L, 33L, 
27L, 29L, 31L, 29L, 26L, 29L, 32L, 20L, 19L, 40L, 42L, 46L, 38L, 
14L, 25L, 24L, 29L, 46L, 45L, 29L, 26L, 34L, 38L, 30L, 33L, 71L, 
52L, 51L, 29L, 33L, 50L, 55L, 40L, 39L, 32L, 34L, 35L, 28L, 37L, 
37L, 36L, 37L, 29L, 52L, 51L, 50L, 44L, 42L, 30L, 43L, 43L, 41L, 
33L, 46L, 49L, 38L, 52L, 50L, 48L, 49L, 49L, 50L, 40L, 39L, 18L, 
NA, 3L, 31L, 43L, 47L), time_seg1 = c(0, 2, 2, 2, 2, 2, 2, 2, 
2, 2, 2, 2, 0, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 0, 2, 2, 2, 0, 2, 
0, 2, 2, 2, 0, 1, 2, 2, 2, 1, 2, 2, 2, 2, 2, 0, 1, 2, 0, 1, 2, 
2, 0, 1, 2, 2, 2, 2, 2, 0, 1, 2, 2, 0, 1, 2, 2, 0, 1, 2, 2, 2, 
2, 0, 2, 1, 2, 2, 0, 2, 2), time_seg2 = c(0, 0, 2, 3, 4, 1, 2, 
3, 0, 0, 3, 4, 0, 0, 0, 1, 2, 3, 4, 1, 2, 3, 4, 0, 0, 3, 4, 0, 
0, 0, 2, 3, 4, 0, 0, 1, 2, 4, 0, 0, 1, 2, 3, 4, 0, 0, 0, 0, 0, 
0, 1, 0, 0, 0, 1, 2, 3, 4, 0, 0, 1, 3, 0, 0, 2, 4, 0, 0, 1, 2, 
3, 4, 0, 2, 0, 1, 2, 0, 2, 3), ed_dich = structure(c(2L, 2L, 
2L, 2L, 2L, 2L, 2L, 2L, NA, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L, 2L, 
2L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 1L, 1L, 1L), .Label = c("< HS", 
">= HS"), class = "factor")), row.names = c(NA, -80L), class = "data.frame")

question from:https://stackoverflow.com/questions/65602634/plotting-piecewise-growth-curves

Answer 1

What I think you want is a piecewise linear spline. You can do this with a truncated power basis function. In your model, you would include time and a function that is time-2 if time is greater than 2 and 0 otherwise. This makes a piecewise linear function that meet each other at time=2. You can do this in the model as follows:

library(lme4)
mod <- lmer(sdmtwr ~ time + I(ifelse(time > 2, time-2, 0)) + 
              (1 |id), data=tmp, REML=TRUE)

Then, you could use the ggpredict() function from the ggeffects package to produce the plot:

library(ggeffects)
g <- ggpredict(mod, "time")
plot(g)

Note: I couldn't get it to run with random effects on the time variables, but with more data perhaps you'll be able to get it to work.