I have a dataset that I'm trying to fit a piecewise linear regression model using segmented(). I wanted to leave the psi = NA because I have several "transects" of data that require a piecewise model to be fitted. The trouble is that I keep getting one of two errors:
Error in segmented.lm(transect.lm, seg.Z = ~distance, psi = NA, control = seg.control(n.boot = 30, :
only 1 datum in an interval: breakpoint(s) at the boundary or too close each other
Error: at least one coef is NA: breakpoint(s) at the boundary? (possibly with many x-values replicated)
I want to leave psi as NA. Can anything within seg.control be changed to prevent the error?
Here is I put in for the segmented model:
transect.lm <- lm(elevation~length, data = data2)
transect.pw <- segmented(transect.lm,
seg.Z = ~length,
psi = NA,
control = seg.control(
n.boot = 30,
h = 0.001,
stop.if.error = FALSE))
I found a post in R-Bloggers that talked about how to manage errors in for loops. The part of the for loop that was giving me trouble can be within a try() command. Within the article, R would just skip over it and continue on its analysis, but I found with my for loop, it would cycle back and try again, so it ended up looking like this:
transect.lm <- lm(elevation~length, data = data2)
try(transect.pw <- segmented(transect.lm,
seg.Z = ~length,
psi = NA,
control = seg.control(
n.boot = 30,
h = 0.001)))
Errors would still be printed to let me know something wrong happened, but it would loop back and try again until a model was fitted, and then move to the next transect.
Thanks for everyone's help!