0

I managed to come up that it would be good to use normal distribution and standard deviation acquired from it since there's no value to compare with. But I have no idea how to proceed. Is there any known method which uses normal distribution?

i9100
  • 407
  • 1
    I think you need to specify your problem a little more carefully. "how to deal with" is not clear without context. Are you trying to estimate the error on your measurement? – Floris May 29 '18 at 13:01
  • Yes. That's what I wanted to know. I'll take your advice. – i9100 May 29 '18 at 13:02
  • So you basically did an experiment where you got some results, and now you want to know how to calculate the error in them? – Nat May 29 '18 at 13:06
  • If you search "error estimation" on this site, you get 71 hits. Many of those contain valuable information that relates to what you are asking about. – Floris May 29 '18 at 15:20
  • Yes. Especially want a method which uses normal distribution. – i9100 May 29 '18 at 15:35

1 Answers1

0

The typical situation is:

For "manual" measurements, it's good to suppose a uniform distribution, so

$\sigma = \frac{resolution}{\sqrt{12}}$

If the result is not a direct measurement, but a result of using measurements like the one above into a formula, you have to use error propagation.

However, you must take into account all contributions, like systematic errors, offsets, developers' specifications, and so on.

FGSUZ
  • 8,293