0
$\begingroup$

I've been wondering why most publications give stats as mean +/- standard deviation, even for things like measurement devices, where reproducibility is a major concern. (E.g., for a given sample, our devices measured 50 +/- 5). Wouldn't it be be more appropriate to use something like a confidence interval?

Is there any reason why using standard deviation (or RSD) alone is the norm?

Any insight is appreciated, thanks!

Improve this question
$\endgroup$
5
  • 2
    $\begingroup$ Your question is based on the flawed assumption that this is true. Based on my experience, most of the publications use rather confidence intervals or mention standard errors, rather than standard deviations. $\endgroup$
    – Tim
    Commented May 18, 2023 at 12:35
  • 1
    $\begingroup$ I have seen mean plus/minus standard error or two standard errors (which is a rough confidence interval). I'm not so sure that standard deviation is "the norm". Note also that the standard error is the standard deviation of the mean, so terminology may be confused. Other than that, the standard deviation indicates the variation of the observations (rather than of the mean itself), which is often of interest. $\endgroup$
    – Christian Hennig
    Commented May 18, 2023 at 12:38
  • $\begingroup$ Thanks for the replies. To clarify, I am looking at sensors, where I've commonly seen the calibration curve presented as the mean +/- st. dev. of 3 seperate sensors. Some do use confidence intervals, but others just use standard deviation. $\endgroup$
    – user388264
    Commented May 18, 2023 at 12:58
  • $\begingroup$ It depends on the purpose of the report: sometimes standard deviations are more appropriate (for calibration and prediction, for instance) and other times standard errors or margins of error are needed (for inference). So I have to agree with @Tim: this question relies on a false premise. $\endgroup$
    – whuber ♦
    Commented May 18, 2023 at 13:31
  • $\begingroup$ FWIW, this is extremely common in psychology. It probably heavily depends on the scientific field. And yes, I also wonder. $\endgroup$
    – Stephan Kolassa
    Commented May 18, 2023 at 13:49

0

Reset to default