How can I avoid uniroot error that stops the loop?

Question

I am running uniroot function in the loop, but hit an error and the code stopped. The code is as below;

func <-function(f) -a*b/c*0.5*d*e^2 + (d/f-1)*g*sin(h*(pi/180))-i
dat <- data.frame(a = c(0.99,0.99,0.99),
                  b = c(0.1986572,0.1986572,0.1986572),
                  c = c(237.5,237.5,237.5),
                  d = c(1028.372, 1028.711, 1028.372),
                  e = c(2.46261, 2.986461, 2.46261),
                  f = c(-1,-1,-1),
                  g = c(9.8,9.8,9.8),
                  h = c(-54.97964, -51.65978, -54.97964),
                  i = c(0.03699588, -0.0375189, 0.03699588))

for(j in 1:length(dat$a)){
   a <- dat$a[j]
   b <- dat$b[j]
   c <- dat$c[j]
   d <- dat$d[j]
   e <- dat$e[j]
   #f: this should be solved by uniroot
   g <- dat$g[j]
   h <- dat$h[j]
   i <- dat$i[j]
   sol <- uniroot(func,c(0, 2000),extendInt = "yes") 
   dat$f[j] <- sol$root
   print(j)
}

Running above code, hit the below error:

[1] 1
Error in uniroot(func, c(0, 2000), extendInt = "yes") : 
      no sign change found in 1000 iterations

The code stopped at j=1, and did not go to j=2 & 3. Therefore, dat$f shows

> dat$f
[1] 1526.566   -1.000   -1.000

My goal is when uniroot hits an error in a given j, put NA in dat$f[j], and continue the loop by the end.

If this works, dat$f[1] and dat$f[3] should have the same value (=1526.566) using the the above dataframe.

Please advise me on how to deal with the uniroot error.

Answer 1

The code in the question would work if the lower bound is set to 1 instead of 0. The problem is that if f is 0 then func is undefined since f is in the denominator.

While that is sufficient it is suggested you make these changes:

1. Use dat shown in the Note at the end ensuring that it does NOT include f.
2. define func more compactly as shown
3. use try to allow uniroot to continue to run even if there are errors (although there are none in t his example)
4. use a lower bound of 1 since func is undefined at 0
5. pass the jth row of dat to func using an argument of uniroot
6. place the results in res (or NA if that iteration fails) so there is no confusion as to what is input and what is output.

Use the more compact form of func shown and put the results in res.

func <- function(f, data) with(data, -a*b/c*0.5*d*e^2 + (d/f-1)*g*sin(h*(pi/180))-i)
nr <- nrow(dat)
res <- numeric(nr)
for(j in 1:nr){
   sol <- try(uniroot(func, c(1, 2000), data = dat[j, ], extendInt = "yes") )
   res[j] <- if (inherits(sol, "try-error")) NA else sol$root
   print(j)
}
## [1] 1
## [1] 2
## [1] 3
res
## [1] 1526.566 2014.476 1526.566

Note

dat <- data.frame(a = c(0.99,0.99,0.99),
                  b = c(0.1986572,0.1986572,0.1986572),
                  c = c(237.5,237.5,237.5),
                  d = c(1028.372, 1028.711, 1028.372),
                  e = c(2.46261, 2.986461, 2.46261),
                  g = c(9.8,9.8,9.8),
                  h = c(-54.97964, -51.65978, -54.97964),
                  i = c(0.03699588, -0.0375189, 0.03699588))