Panic when computing a confidence interval from a single sample.

[vext01]

Hi,

We use your library (thank you for writing it) to compute confidence intervals for benchmarking data.

Our continuous integration pipeline does a "quick run" of our benchmarks and runs our statistics program to check that it doesn't obviously bork. A "quick run" here runs each benchmark only once, which means that we end up computing confidence intervals from a single sample and this crashes stats-ci with a panic.

Minimal example:

use stats_ci::{mean, Confidence};

fn main() {
    // 2 samples: this is OK.
    let ci = mean::Arithmetic::ci(Confidence::new(0.95), &[1.0, 1.0]).unwrap();
    dbg!(ci);

    // 1 sample: this crashes:
    // thread 'main' panicked at .../stats-ci-0.1.1/src/stats.rs:34:64:
    //   called `Result::unwrap()` on an `Err` value: BadParams
    let ci = mean::Arithmetic::ci(Confidence::new(0.95), &[1.0]).unwrap();
    dbg!(ci);
}

The panic comes from an unwrap in stats-ci.

Obviously it's kind of nuts to be computing intervals for a single sample, but at the same time, I don't think this should cause the library to panic.

What should happen though?

• Compute an interval for the single sample, if possible?
• If not, the library should probably raise an error to the caller rather than panicking internally.

I prefer the former, but appreciate that it may not be possible. What do you think?