How can we correctly calculate the following?
df <- data.frame(logp=c(-2,-200,-600))
10^(df$logp)
# [1] 1e-02 1e-200 0e+00
This will give us 0 in logp=-600. Can we handle this problem?
Asked
Active
Viewed 42 times
1
user20650
- 24,654
- 5
- 56
- 91
user224050
- 317
- 3
- 10
-
Perhaps you could calculate it on the log scale (`log(10)* df$logp`) or use a high precision library (`10^Rmpfr::mpfr(df$logp, 120`) -- depends on what you want to do with it. – user20650 Jan 09 '21 at 16:13
-
The function `.Machine` provides a list of numerical characteristics and the smallest non-zero normalized floating-point number ~1E-308 – Dave2e Jan 09 '21 at 16:16
-
?duplicate: https://stackoverflow.com/questions/5802592/dealing-with-very-small-numbers-in-r – user20650 Jan 09 '21 at 16:38