Search All of the Math Forum:

Views expressed in these public forums are not endorsed by NCTM or The Math Forum.

Notice: We are no longer accepting new posts, but the forums will continue to be readable.

Topic: NHST : Accuracy and Precision
Replies: 0

 Luis A. Afonso Posts: 4,758 From: LIsbon (Portugal) Registered: 2/16/05
NHST : Accuracy and Precision
Posted: Oct 14, 2013 4:49 PM

NHST : Accuracy and Precision

http://en.wikipedia.org/wiki/Accuracy_and_precision

In classical Null Hypotheses Statistical Tests the ONLY result/conclusion we can attain, given data and confidence level:
___we obtain sufficient evidence the null is unlike
___we does not.

Any other one more *specific* sometimes found in text-books /papers is unlawful/absurd/wrong.

The difference of two means

The true difference between Population means. D= muX- muY is accurate (ACCURACY) relative to the observed difference, D´= xbar-ybar, if | D - D´ | is short. By the other hand high PRECISION demands that the Confidence Interval CI = [u, v] containing D´ with a probability alpha ( = 0.95, so) be short.

_______________________u________v__
___________|____________|____|____|_________
__________ D_______________D´______

ACCURACY and PRECISION have rather different features and controlling ways:
____ACCURACY: we can establish the probability to find D inside the confidence interval, u<= D <=v, say 95% Confidence Level. In contrast PRECISION - how wide or narrow is the interval - is left to the Goods of Chance to decide: we are unable a priori to preview because are dispersion data immediately dependent. It can happens even that a high precise value be be useless/trash because the CI does not contain D . . . a present type I error (fee figure above).

Luis A. Afonso